Cloud robotics is a new method for creating self-governing robots that use “the brain” of the robot as a combination of on-board and remote/distributed computing in the form of cloud-based services. Traditionally, robots have been designed with the onboard computer performing all functions related to the robot. Cloud robotics enables robots to access high-performance servers (usually remotely) over the internet to process sensor data, learn from experience, and access shared knowledge and resources. By using cloud robotics, users could realize cost savings, additional functionality, and easier software upgrades for their robotic platforms.
The ability of cloud robotics to increase the processing capacity available to robots is another advantage of this technology. The majority of robots will need to perform complex operations such as speech recognition, object identification, and safety planning, most of which require substantial processing power. Because these tasks can be processed by remote servers in the cloud, there is no need for the robots to include large processing capacity, thus making the overall design and operation of the robots less expensive.
Cloud Robotics is a new technology trend for creating autonomous systems. Autonomous systems, in this case, will have their “brain” (all functions of the system) created through remote and/or distributed computing via cloud-based services. Historically, the onboard computer was responsible for all functions of the robot. This has changed with Cloud Robotics. Cloud robotics utilizes high-performance servers (usually located remotely) that robots access via the Internet to process sensor data, learn from experiences, and share knowledge/resources.
This also means that Cloud Robotics offers the advantages of lower costs, greater capabilities, and easier updates to robotic platform software. Another advantage of Cloud Robotics is the increased processing ability. Many times, robots must perform complex tasks such as object recognition, speech recognition, and safety planning. These tasks usually require a large amount of processing power to execute. By executing these tasks in the cloud, robots will not need large amounts of processing power, and thus will be less expensive.
#Virtual Robot Testing: Smart & Powerful Ways to Save Time and Money
The hybrid model leverages both the robot’s local processing capabilities (e.g., emergency stop procedures) and the cloud’s large-scale processing capabilities (e.g., advanced computation and long-term machine learning). Another concern for hybrid models is the security and privacy of the data processed by the robot, i.e., whether the data is video-related, location-based, or workplace-related.
With the expansion of 5G networks and the advancement of AI, we can expect an increasing number of applications that use cloud robotics across many areas of consumer life, including homes, hospitals, retail, agriculture, and delivery. Cloud Robotics involves integrating a capable local robotic system with scalable cloud-based technologies. When successfully deployed, cloud robotics enables multiple robots to communicate and coordinate in real time, learn faster than traditional robots, and ultimately deliver more value to consumers on a daily basis.
Have you ever seen your robot vacuum sucked into that same chair leg for the tenth time and wondered why it didn’t have a better brain? Well, the problem is that the robot’s brain is locked inside its body. But what if this robot could use an internet-supplied super-brain instead of just the computer onboard?
To understand how to connect your smartphone to the internet-powered cloud, think about your phone. It can take a picture on its own, but the real power behind the camera lies in the cloud’s massive processing power. Google Photos uses cloud-based processing power to scan all your photos and detect faces, a task far too complex for your phone alone to handle.
Using Cloud Robotics the same way: Once the robot views the object through its built-in sensors, it transmits that information to the cloud, where complex processes are performed on the information; the robot then receives a simple instruction, like “pick up the red ball.”
Cloud Robotics allows a machine to perform actions in the physical world while providing remote, powerful “thinking” capabilities.
Ultimately, the robot will act as the physical embodiment of the cloud’s vast intelligence, enabling powerful cloud computing for robotic learning so that each robot can use its own learning experiences, which are applied instantly to all other robots.
This model has also allowed for the emergence of new ideas, such as Robot as a Service (RaaS), and has turned the cost-prohibitive process of buying intelligent machines into an affordable subscription service.
Cloud Robotics Market and Adoption Statistics
Source: MarketsandMarkets Cloud Robotics Market Report
https://www.marketsandmarkets.com/Market-Reports/cloud-robotics-market-198734864.html
Cloud Computing: Foundation enabling scalable robotic intelligence
Cloud computing allows today’s robot systems to scale from a single unit to multiple units connected, enabling them to learn and operate at scale. Rather than each robot bearing the costs of heavy AI processing hardware, Cloud Computing enables robots to tap into scalable resources (i.e., processing power, storage, and other services) on demand. As such, Cloud Robotics is being utilized across warehouses, hospitals, and smart cities, where many robots are expected to operate reliably and simultaneously.
Cloud computing allows robots to “outsource” computationally complex workloads, including vision-based identification, mapping and large-scale model prediction to remote servers in the cloud. Teams can use identical perception pipelines across thousands of robots running the same models, and the ability to rapidly update these models provides efficiency during performance upgrades and reduces equipment costs.
Additionally, Cloud computing will increase “Fleet Intelligence”. Data from sensors used by robots is voluminous, and the Cloud’s capabilities to collect, label, and analyze this data to identify and resolve “edge case” issues and improve robot behavior will be enhanced. Improvements to a navigation policy, for instance, learned during the operation of one robot, can be directly applied to every other robot in the fleet operating in the cloud.
#Cobots for Small Businesses: Powerful & Practical Why It Matters
Finally, Cloud Computing can facilitate very large-scale simulation. With this capability, teams can now simulate new software in virtual environments before deploying it on actual “production” machines, resulting in improved efficiency and reduced risk of machine downtime.
The primary advantages of using Cloud Computing for Robotics relate to operational capabilities. For example, Robot Health Monitoring, Log Analysis, and Alerting help operators identify potential problems in their robotic operations and schedule regular maintenance; these are just a couple of the many types of continuous processes that can run as long as you have access to the Cloud via Computing.
Additionally, Cloud Computing enables Remote Access, Role-Based Permissions, and Audit Trails, which will become increasingly necessary to ensure regulatory compliance in environments such as Healthcare, where your cloud robotics application is being deployed.
While Cloud Computing offers numerous benefits for Robotics, it also has several limitations. To offload successfully, robots need reliable network connections; otherwise, high latency will limit what can be offloaded to the Cloud. For Safe operation of robots, Cloud Robotics Systems typically maintain Real-time control and emergency behaviors at the local robot level, using the Cloud for higher-level planning, learning, and coordination. As a result, it is imperative that cloud computing platforms are secure, as they are responsible for protecting sensitive data such as video, location, and facility information.
To summarize, cloud computing provides the scalable infrastructure needed to transform Robots into continually improving systems, and cloud robotics offers an operational approach for delivering this infrastructure to real-world machines.
AI Robotics: Combining artificial intelligence with robotics
Traditional robotic programming involves setting specific predefined rules or “scripts” that govern how a robot responds to its environment. In contrast, AI Robotics provides an opportunity for machines to take input from their surroundings (for example, what they see) and anticipate the effects of possible actions before taking them. As modern robotic systems become more connected to cloud robotics, robots can share data and use cloud-based computing services when necessary.
One of the main goals of AI Robotics is perception, converting raw sensor input into usable information. Computer Vision is one area of AI Robotics where systems can detect and locate objects, determine distance, and track movement; Machine Learning enables AI Robotics to recognize objects under varying lighting conditions, object orientations, etc. When using Cloud Robotics, the system can upload frames/images to the cloud for processing at greater intensity and/or model updates, while still maintaining critical safety features on the local device.
A second area of interest in AI Robotics is planning and control – how a robot determines which path to take through a particular environment and how it will interact with objects in that space. For example, the robot will have to find a safe route around obstacles and adjust its grip on objects as necessary (for instance, when objects fall or are jostled). Reinforcement Learning and Imitation Learning in AI Robotics can also contribute to the ongoing improvement of a robot’s capability to develop and manage its own movement planning and control.
Another area of interest is human-robot interaction. As AI Robotics provides capabilities for speech recognition, gesture recognition, and/or intention detection, robots may be able to interact with humans in more natural ways. In operational environments, robots can be designed to slow their pace near a person, understand and respond to verbal commands or requests, and perform well even when working from established routines. One benefit of using Cloud Robotics to facilitate this kind of collaborative effort is the potential for the remote sharing of maps, task libraries, and/or language models, which would enhance consistency of communication across multiple robots.
The greatest benefit of AI Robotics to teams will likely occur with a hybrid model (i.e., real-time response/reaction/emergency behavior at the local level vs. training, evaluation, etc., at the remote level). Cloud Robotics supports the remote aspects of the hybrid model (monitoring, updating, shared learning) while avoiding any risks to the safety of the robot/local personnel/remote personnel. The further development of AI robotics in logistics, healthcare, agriculture, and home assistant applications relies upon scalable architectures (such as those provided by cloud robotics).
Robotics Technology: Core technologies shaping modern robotics
“Robotic technology includes both the physical hardware technologies used in modern robots (e.g., sensors and motors/actuators) and the software (operating systems, programming languages, and artificial intelligence) used to create robots. Unlike traditional automated manufacturing, which relies heavily on mechanical means, robotic technology has evolved by combining mechanical design, electronic controls, intelligent control systems, and sophisticated algorithms to enable robots to safely navigate a changing environment, be aware of their surroundings, and complete meaningful tasks. Today, many robots use cloud-based robotics for rapid scaling and enhanced functionality.
Robotics technology also includes several types of hardware, specifically actuators (motors, servos, hydraulic cylinders), transmissions, and end-effectors (vacuum grippers, tooling, etc.). The selection of these hardware components directly influences a robot’s ability to manipulate objects, maintain precision, and endure the rigors of daily operation. Additionally, robotics technology uses numerous different types of sensors to provide a robot with feedback about how it is operating in an uncertain environment.
For example, camera sensors, depth sensors, LiDAR, IMU sensors, force torque sensors, and encoders all help a robot understand its surroundings. Utilizing cloud-based robotics can assist with the aforementioned hardware and sensor components by allowing users to access calibration tools remotely, manage configurations from a central location, and monitor entire fleets of robots.”
#What Is a Digital Twin? 5 Powerful Ways Virtual Systems Create Smart Value
Robotics software technology is directed toward addressing issues relating to real-time control, motion planning, and safety systems. Robot controllers will translate higher-level objectives into smooth, safe motion plans that account for the robot’s limitations to prevent damage or injury to humans and equipment.
Additionally, the development of robotics technology included the incorporation of localization and mapping, which enable robots to understand their position and navigate their surroundings. Cloud robotics technology may assist in this area by providing access to shared maps for all robots, directing traffic flow across multiple robots, and handling computationally intensive tasks when local hardware lacks sufficient capability.
The emergence of AI and Machine Learning has fundamentally altered how we view and utilize robotics technology. For example, vision models enable robots to identify objects, determine the object’s orientation, and detect flaws. Also, the use of learning-based methods to grasp and manipulate objects improves performance, as these systems can adapt to a wide range of objects.
Cloud robotics enables multiple robots to collect data simultaneously, which can then be used to train models on the same dataset and update them in real time for the robots. This creates much shorter development cycles than updating one robot at a time.
Robot Operation tasks are being expanded through connectivity to include other critical functions, such as security through communications, authentication, logging, and update capability, to allow robots to function securely. Cloud robotics also provides remote diagnostics, performance monitoring, and quick troubleshooting functions through secure connectivity.
Robotics Technology must also be reliable; it is essential that critical functions such as obstacle detection/avoidance or emergency stop continue to operate when connectivity is lost.
Robotics Technology includes hardware, software, and intelligence to enable today’s robots to perform their tasks, and cloud robotics technology combines these elements in an integrated way to create scalable and share learning operational systems.
Core Technologies Behind Cloud Robotics
Source: IEEE Robotics and Automation Society
https://www.ieee-ras.org
Intelligent Robotics: Robots capable of learning and decision-making
Intelligent Robotics is a type of robot that learns from its experience, and it uses this experience to decide what it wants to do and how it should react to an ever-changing environment. In contrast to traditional “dumb” robots that run through a set sequence of actions and do not have the capacity to adjust their behavior based on changes in the environment (for example, if the shape of the objects changed, or there was a blockage in the path) intelligent robots have the capability to adapt to changes in the environment.
A number of applications also leverage cloud robotics to extend this intelligence to many robots.
Perception/Understanding and Learning are two of the primary characteristics of Intelligent Robotics systems.
In terms of Perception/Understanding, Intelligent Robotics systems take input from various devices (such as cameras and depth sensors) and translate it into useful information, such as identifying objects, determining locations, and detecting hazards. Some Intelligent Robotics systems may also utilize learned models to represent the unpredictability of the real world.
Cloud Robotics enables centralized development of complex perceptual models and distributes them to each robotic unit, so that any changes to a model are automatically propagated throughout the entire fleet.
Intelligent Robotics Systems utilize supervised learning, reinforcement learning, and/or imitation learning to improve the behavior of their robotic agents (i.e., grasping, sorting, etc.) in addition to using cloud-based intelligent robotics systems to enable shared learning through the utilization of logged data from one robotic agent to improve/train other robotic agents performing the same or similar functions. Warehouse/hospital-type applications are particularly well-suited to this method as they have large numbers of robotic agents performing similar tasks.
Typically, the decision-making process used by Intelligent Robotics Systems involves trade-offs among conflicting objectives (e.g., fast vs safe; efficient vs battery life; completing tasks vs providing users with a positive experience) regarding which objective(s) take precedence. One way to apply intelligent robotics is to use Predictive Outcome Planners to select robot actions that minimize potential risks. Cloud Robotics can assist in planning and executing tasks by distributing maps, rapidly assigning tasks to robotic agents, and ensuring coordination amongst multiple robotic agents to avoid bottlenecks and promote collaboration.
Cloud robotics has both positive and negative implications regarding reliability and performance (in addition to the intelligence) in a cloud-enabled robot system. The system will need to be monitored and tested, and can be updated quickly. A cloud robotics system can support the monitoring and testing of a robot through log collection to track trends and patterns; anomaly detection to identify when something is going wrong; and remote diagnostic capabilities to minimize downtime and maintain high-performance operation.
However, using cloud robotics introduces additional network and security concerns, which have led many organizations to retain local control over safety-related functionality while using the cloud for training, analytics, and coordination.
Thus, intelligent robotic systems are a class of robots capable of continuously learning from their environment and making decisions in response. Conversely, cloud robotics represents a level of abstraction that provides scalable, dynamic, and accelerated learning for intelligent robotics systems.
Cloud Automation: Automated Processes Powered by Cloud Systems
“Cloud automation allows you to automatically run your cloud-based applications, eliminating repetitive user actions, including launching new jobs, transferring data files, and manually integrating systems. Cloud automation also allows organizations to connect applications, data sets, and infrastructure, and to trigger automated processes based on specific events, times, or rule-based criteria (e.g., “when a sensor alerts us, generate a ticket and notify the on-call team”). Organizations that implement cloud automation can increase their operational speed, consistency, and scalability.
The major components of cloud automation are triggers, workflows, and managed services. As a result, organizations that use cloud automation can dynamically add compute resources as needed, send messages between systems, and delete unnecessary resources. Cloud automation also supports the concept of “Automation as Code” – a method for defining repeatable steps to deploy applications, back them up, set up environments, and perform other common tasks. This concept has two benefits for organizations. First, it eliminates the risk of manual errors; second, it greatly improves the management of complex systems as they grow.”
Using cloud automation in conjunction with cloud robotics offers the best opportunities to leverage both for robotics and physical activities. In cloud robotics environments, a cluster of robots can generate large volumes of log, image, and telemetry data. Cloud automation can be applied to collect and store all data generated by robots and run all necessary data analysis pipelines. Cloud automation can also schedule software updates, roll back problematic releases, and push new configurations to a large number of machines in cloud robotics environments with little to no downtime.
By applying cloud automation, you can achieve greater reliability by enabling it to monitor and respond to issues. If an issue is detected, an automated diagnostic can be triggered, a service restarted if it is not running properly, or workloads redirected to working portions of the environment. In a cloud robotics environment, an operator can quickly determine where problems are occurring within their fleet(s) and apply standard “playbooks” to fix them, rather than applying “band-aid” fixes to every problem they find.
Eventually, cloud automation can help your team develop self-healing systems that automatically recover from common failures.
Cloud Automation can enforce usage policies, rotate credentials, encrypt data, and automatically create logs. Therefore, when using robots in cloud environments for highly sensitive applications (e.g., healthcare) or for streaming data, implementing automation controls with Cloud Automation helps reduce the organization’s risk while allowing it to continue operating without disruption. Also, using Cloud Automation to perform compliance-related functions, such as maintaining data retention records and reporting incidents, automates many administrative tasks.
Cloud Automation has enabled organizations to convert difficult-to-use, multi-step processes into consistent, repeatable workflows, while Cloud Robotics applies those workflows to real-world objects via connected devices. Together, Cloud Automation and Cloud Robotics enable organizations to grow more rapidly, respond more quickly to disruptions, and deliver more consistent services across all their locations and devices.
Remote Robotics: Robots operated and managed through the cloud
Remote Robotics is the use of cloud services to control, manage, and update remote robots via the internet. In Remote Robotics, a robot will stream back sensory data (video, depth, telemetry) from a local location to operators located at a dashboard while the operators send forward commands for navigation, inspection, or task execution. This model is very closely tied to cloud robotics, where cloud infrastructure provides scalable compute, storage, and coordination across many machines.
The common environment for Remote Robotics is “human-in-the-loop” control; the local robot handles safety and stabilization, but the remote operator can take over when things get tricky, confirm decisions, and guide manipulation. Remote Robotics also makes it possible for fleet management – assigning tasks, pushing software updates, and tracking performance from one operation center to dozens or even thousands of devices. In cloud robotics, robots can share maps, receive AI model updates, and use centralized logging to detect issues earlier.
#How AMRs Navigate Warehouses and Hospitals: A Smart Essential Guide
Remote robotics provides great value in situations that are either dangerous, remote, or difficult to staff with humans. Examples of this include inspecting industrial facilities, monitoring warehouses after business hours, inspecting infrastructure, conducting disaster response efforts, and assisting with healthcare logistics. In these situations, remote robotics reduces the need for travel and allows specialists to assist multiple locations. Cloud robotics makes this possible by providing identity management, secure access controls, and standardized data pipelines between robots and the cloud, so that all robots report their status and receive instructions in a standardized format.
There are also some real-world limitations. Remote robotics depends on a reliable connection to operate the robot; latency can affect how well you can control your robotic system from a distant location. To ensure safety during deployments, many remote robotics solutions use a hybrid approach. The robot has local capabilities (emergency shutdown, collision detection, and basic control), but high-level planning and analysis occur in the cloud.
Cloud Robotics will help secure remote robotics through data encryption, user authentication and authorization, and access controls. The combination of automation for failure over and constant monitoring of robotic systems enables redundant and tolerant robotic systems in a cloud environment when implemented correctly.
Overall, Remote Robotics offers remotely operated robotic services, and Cloud Robotics provides a scalable and secure base layer on which Remote Robotics can operate.
The Hive Mind: How One Robot’s Lesson Instantly Teaches Thousands
One of the largest benefits of Cloud Robotics is the ability for a “Group” to learn together. When you are using a GPS like Google Maps or Waze to get directions from work to your home, if even a few users in that area slow down, the program will update all GPS users’ routes to help them avoid traffic congestion. The GPS device on each user’s car did not actually report traffic congestion; rather, the App running on each user’s smartphone reported congestion based on other users’ reports. That is exactly how large fleets of robots will become “smarter”.
Now, picture a large warehouse with 1,000 robots, all connected. One of the robots grabs something, and after trying many ways to grab it, the robot is able to find a way to effectively pick up the object (grip). The robot then saves this “method” for grabbing the object to the cloud brain (or shared knowledge base), and as soon as that is done, all 999 other robots in the system are automatically given the updated information about how to pick up the same type of object.
Since every robot has access to the current information being used by all other robots in the network, the group’s total ability to solve problems grows exponentially. If one robot encounters a problem it has never seen before, all other robots in the network have immediate access to the solution developed by that robot; therefore, the group can develop its problem-solving abilities faster than any single robot could if it were developing its own. Since the group intelligence exists in the cloud, each individual robot can be designed to do much less than a human and still contribute to the group intelligence.
How the “Hive Mind” Learning System Works
Example: Autonomous Warehouse robots continuously share navigation improvements across entire fleets
Source: Google Cloud Robotics Platform
https://cloud.google.com/robotics
Why Cloud-Connected Robots Can Be Lighter, Cheaper, and Simpler
The fundamental concept of Cloud robotics is that it is changing how we build robots from a mechanical/physical perspective. Instead of having the computer (the “brain”) as part of the robot itself, the robot’s body becomes just another interface (eyes, arms, etc.) to connect to the Internet, with the “brain” doing the thinking in the cloud.
For example, think of a high-cost laptop ($ 1,000) and a low-cost Chromebook ($200- $ 300). One processes locally; one sends the work to be processed in the cloud. This will likely become increasingly prevalent as more services move online.
This shift toward more affordable “smart” products and cloud-based “brain power” for sophisticated automation represents a revolutionary idea. Now, large corporations are not the only organizations capable of utilizing advanced automation. For instance, smaller warehouses and farms can now rent the “intelligent” capability needed to run their systems at significantly lower cost by using fleets of less expensive, simpler robots.
This also changes who develops robots and how they will be developed. When a robotic system has the “brain” in the cloud, upgrading the “intelligence” of the robot’s brain is as simple as updating the apps on your smartphone.
Upgrading a Robot’s Skills Is Now as Easy as Updating an App
As you can see, when you wake up in the morning and receive a notification that there is an app/system update for your mobile device, you have experienced many of the same benefits that cloud-based robots do. Cloud-based robots are by no means applications or systems; however, both utilize the same fundamental concepts (i.e., receiving immediate updates via the Internet), which have now made it one of the main advantages of utilizing connected robots.
Prior to the development of cloud-based robots, changing the way a robot performs a new task required manual interaction with the physical robot. Service technicians would be forced to physically interact with each machine and reprogram it manually. If a company had a large number of robots (hundreds), the technician would be required to travel to each individual robot to reprogram each one, a time-consuming and expensive process.
Cloud-based learning enables a robot to continually improve its intelligence, efficiency, and capabilities. It is no longer simply a machine but a platform for continuous improvement throughout its lifespan. The machine becomes capable of “constant learning,” which will have a profound impact on the future of robotics and automation. The potential exists to purchase a robot today and, through continuous improvements by the vendor or user, to upgrade and enhance its capabilities for years to come.
For example, a delivery robot may see a new type of crosswalk and report back to the vendor’s cloud server. The vendor then develops a patch to address the issue and deploys it to the robot via the cloud. Once deployed, all other robots in the city are updated to the new version, and each can successfully navigate a crosswalk it has never seen before.
Cloud Robotics in the Real World: From Smart Farms to Your Front Door
Cloud-connected robots are being used in many businesses worldwide and are changing how they operate today. The next time you receive your breakfast, what process did it go through before you received it? Daily operations such as food preparation, mail delivery, and other day-to-day tasks are becoming more efficient with cloud-connected robotics, which can improve quality of life through enhanced sustainability, responsiveness, and efficiency.
Although there are several applications of Cloud Robotics, most involve continuous learning and coordination between the two components.
Three major areas where Cloud Robotics applications are making a large difference include:
- Precision agriculture: A shared, constantly updated image database that smart tractors and drones can use will allow them to distinguish weeds from healthy plants. Smart tractors and drones can spray only a small amount of herbicide directly onto weeds (rather than spraying all the plants), conserve water, and reduce chemical use by using a cloud-based database.
- Warehouse Logistics: In large-scale warehouse logistics systems, thousands of robots can work with each other, communicate about their movement plans, and execute those plans with extreme speed via the cloud. When a new product is introduced into a warehouse, one of the robots will develop an optimal method for gripping it; that method will automatically become available to every other robot in the warehouse.
- Autonomous Mobility: Autonomous cars and autonomous delivery robots need the cloud to access real-time maps of road hazards, traffic, and road construction. If a vehicle hits a road hazard (e.g., a pothole) on its route, it will identify the hazard’s location and send a warning signal to every other vehicle on the same network, so they may avoid it.
When we connect our robots to the same network — sometimes referred to as the “Internet of Things” (IoT) — it creates a strong loop. Each robot becomes a sensor collecting data and contributes to an overall intelligent pool of information, making the entire system smarter. A robot on a farm in California can help the robot in Florida; a delivery robot in Boston can assist the robot in San Francisco.
This heavy dependence on the constant availability of a wireless connection brings up a crucial concern. What happens to a robot when it loses its Wi-Fi signal while traveling down a busy street or through a crowded warehouse?
Cost Advantages of Cloud-Connected Robots
Example: Cloud robotics reduces hardware complexity and allows robots to rely on powerful remote computing resources
Source: MIT Cloud Robotics research
https://robotics.mit.edu
What Happens When the Wi-Fi Drops? The Biggest Hurdles for Cloud Robotics
In summary, the question of what happens when the internet connection goes down gets to the heart of two major engineering problems in cloud robotics — speed and security. While the potential of cloud robotics is enormous, connecting robots to a remote brain poses numerous engineering challenges.
The first and biggest challenge is communication delay (latency). This is the same delay you experience while playing an online video game or during an international video call, and it could be catastrophic for a robot like a self-driving car. For example, if a robot has to wait less than a second for a cloud-based system to send it a “brake” command, it may not have enough time to react and avoid an accident. The need for a robot to respond quickly will be a basic constraint on its use.
In addition to connection speeds, there is the matter of security. Any computer connected to the internet or smart speaker at home, for example, is vulnerable to hacking; so too can a robot with a cloud connection. A hacker could accident or damage. Furthermore, if the robot has been hacked, it can be used to monitor people’s activities, and hackers can violate their privacy by stealing sensitive video footage from a production line or inside someone’s home. The protection of that network is thus no less than the need to increase the speed of that network.
Because of the challenges associated with this type of issue, designers face a major decision right now. If we can’t guarantee constant, fast, and completely secure connections, what level of “thinking” can/should occur on the robot itself versus on the cloud? This thinking process is currently creating a new discussion: where should a robot’s brain be (the cloud) vs. where should the brain be placed (locally – “edge”)?.
Real-World Cloud Robotics Applications
Example: Companies like Amazon, JD.com, and Ocado operate thousands of cloud-coordinated warehouse robots.
Source: McKinsey Robotics Industry Report
https://www.mckinsey.com
Cloud vs. Edge: Where Should a Robot’s Brain Live?
There are two kinds of cloud brains (or what I will call “brain”, one near, the other far), which have developed an important new technology called Edge Computing. The term “edge” indicates that the processing occurs where the data is being generated – i.e., on the robot itself or as close to the end of the Internet as possible.
You can think of this concept as the speed at which you pull your hand away from a hot stove. Your hand leaves the stove almost immediately, before you even realize that it left the stove – there is no need to think about moving your hand, it happens automatically. Similarly, the way a robot works is just like Edge Computing – the robot gets the same instant reaction time for all of its critical functions that require no delay.
A robot’s onboard computer is well-suited for instantaneous reactions; however, it lacks sufficient raw processing power to perform large-scale computations. On the other hand, the cloud is for big picture stuff. For example, a self-driving car might use its edge computer to instantly slam on the brakes if it sees someone crossing the road, but it uses the massive computational resources provided by the cloud to analyze traffic patterns across the entire city to find the best route home.
In real life, the most advanced systems do not provide a forced choice; they use a Hybrid Computing System. In essence, this is the “best of both worlds” scenario. The robot has a local (and very fast) brain for performing all of its core functions and immediate survival needs, while remaining connected to a much greater super-brain called the Cloud to perform other functions, such as learning new skills, making judgments about very complex issues, etc., as well as to receive periodic upgrades.
The ability to respond quickly and independently, and to continually learn and improve through experience while working within a larger network, is what will unlock the next major breakthrough in hybrid computing systems. This ability to work both as a single unit and as part of a larger group will fundamentally alter how we interact and work with each other over the next few years.
Our Shared Robotic Future: How a Connected World Gets Work Done
The era of independent robots has come to an end. Cloud Robotics shifts the focus away from wires & metal and into a large network that gives robots life – a great shift in direction that allows robots to be linked to a central intelligence and to learn from each other instantaneously. Understanding the benefits of shared intelligence is what will lead us to the future of automation.
Cloud Robotics is changing how we use our machines. It is the beginning of eldercare robots learning new forms of assistive technology from one another, agricultural robots learning to prevent food spoilage, and rescue drones working together as a team in a disaster area. Cloud Robotics is not an incremental update – it is laying the foundation for technology that works with and for humans in ways we have not even thought of yet.
A self-driving car or a “Smart” Warehouse is more than a machine – it is a student in a global classroom connected by the Internet. Just as the Internet connected people to share knowledge, Cloud Robotics connects machines to share skills. We are about to develop a Global Physical Intelligence that will forever change how we work and live.
Comments 3