Industrial robotics has been around for decades and has transformed the industry. Well-established high-precision engineering with industrial robotic arms fuels today’s factories, and large factory jobs are fully automated. This will continue to pick up pace.Content Continues Below
Practical usage of mobile robotics started its journey in industrial environments with automated guided vehicles, which were also the first step to making mobile robots autonomous. But they are restricted typically to factory and shop floors operating on a known path and using external infrastructure. They can also use sensors, but are restricted in types and run on same paths. They detect obstacles and can stop, but do not navigate around them.
The last few years have seen astonishing progress in sensor technologies, available compute power, edge sensor intelligence/AI and cloud, as well as dramatic cost reductions.
All this puts the autonomous mobile robots (AMR) segment on the cusp of taking off. AMR can operate on various terrains, sense, plan, navigate, detect and avoid obstacles. They can be reconfigured for newer terrains and newer tasks.
This article is the first among a series of articles on AMRs and in general rapidly growing field of mobile robotics.
Autonomous mobile robots
There are various categories of mobile robots, including land-based robots, such as self-driving cars, trucks, industrial mobile robots, surveillance and service robots; aerial robots, including unmanned aerial vehicles and drones; and underwater robots. When intelligence is added to these robots to allow them to plan their paths and navigate autonomously, the same category of robots represents the world of autonomous mobile robots.
AMRs have a wide range of applications, including industrial, service robots, surveying, mapping, surveillance, agriculture, healthcare, defense, space and more.
When dirty, dangerous and dull jobs come into the picture, so do robots. This article gives a good summary.
While self-driving cars, robo taxis and trucks, a category of AMRs, hog the limelight today, there is a silent revolution happening in various other AMR segments for a number of industrial and service automation categories.
Autonomous operation can be partial or full, but in each case there is invariably a need for manual intervention in emergency situations or when a robot can’t handle fully autonomous operation. Hence, the more relevant category of robots is controllable autonomous mobile robots, which always have manual controls for an operator to override fully autonomous operations.
Why is the AMR industry on an inflection point
The smartphone revolution did three important things to the industry. First, it made rapid advances in sensors that can be used for commercial applications. It likewise dramatically reduced sensor prices and also had an indirect effect on other sensors, which are not necessarily part of a smartphone, both in terms of rapid technological advances and dramatic cost reductions. Second, compute power both on the edge and cloud exploded. Third, AI, machine learning, deep learning and advanced sensor fusion techniques took off, again benefiting from large compute power and commercially viable costs.
Another revolution happened during the same period which changed the entire software industry: open source software and its community. This put complex but siloed technologies out in the open for collaboration and wide usage. The same thing is happening in the robot world with open source community-driven efforts such as Robot Operating System (ROS).
Combining all these factors, the AMR industry is poised for a rapid takeoff.
In addition, with cloud and IoT, we now have cloud-controlled managed robots with some intelligence and AI also processed on the cloud. This opens up many applications where a fleet of robots can be managed for various operations, their real-time data observed, processed, intelligence gathered and shared across fleets of robots.
One needs to be aware that AMRs going astray through loss of control or failure of sensors and potentially causing damage is a definite possibility. Therefore, controlled AMR is now the major focus area. There has to be enough redundancy in AMRs to be able to control them, including remotely controlled over the cloud, manually and physically.
Key building blocks of AMRs
While each robot can be unique, there are common elements such as sensors, algorithms for path and trajectory planning, object detection and navigation, mapping, control systems, manipulation/arms, grippers, motors/drives, power/batteries and, of course, the entire mechanical structure and body.
Various sensors on AMRs include cameras (mono, stereo), depth sensors (e.g., time of flight, sensors, Lidar, radar, encoders, ultrasonic sensors, force sensors, inertial measurement units and more. Sensor fusion, deep learning and computer vision, algorithms such as SLAM (simultaneous localization and mapping), trajectory planning, feedback control systems, forward kinematics, inverse kinematics, motor control systems and many more are the heart of these complex electro-mechanical systems.
AMRs also need to be managed by human manual intervention for both safety and for situations where AMRs are stuck. Remote management and control via servers and the cloud is going to be critical for controllable AMRs.
The industry is not without its hurdles
There are technology limitations that need to be resolved when using AMRs. One example is battery limitation in terms of total services duration of the robot without a need for recharging.
Safety also remains a primary concern when running AMRs on different terrains and especially where human density is high and where loss of control of an AMR can lead to damage.
Robot intelligence and control algorithms are complex and vary for different use cases based on type of robot, terrain, applications/use cases, environmental conditions to handle, sensor choices and more. Hence extensive fine-tuning, optimizing and testing are still necessary for the environment they operate in.
There is also no common industry agreed-upon compliance framework for certifying AMRs. The industry has to think hard on how to bring out compliance and certification frameworks.
In the next series of articles, I will explore each aspect of AMR, and in general robotics, in more detail.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.