Positioned snuggly between Hannover Messe, Automate in Chicago, and ICRA in Tokyo, WTWH Media’s Robotics Summit & Expo in Boston felt like it was bridging the gap between industry and academia. Many Small and Medium Enterprises (SMEs) were present at the event, showcasing innovations that would benefit stakeholders in nearly every field of robotics. Academics, either as founders of companies, members of advisory boards, or representing their institutions, gave talks on research into state-of-the-art robotics, while maintaining a focus on commercial potential.
Robotics parts manufacturers were present in force, as were (smaller) Systems Integrators (SIs) and automation deployment specialists. Of the bigger players within the SI space, conspicuously, only Dassault Systèmes exhibited—with no sign of Hexagon, Rockwell Automation, or Siemens (although the latter gave a talk on digital twins). Likewise, of the main robotics manufacturers, only KUKA had a permanent show stand; ABB, FANUC, Yaskawa, and Teradyne were not officially present. However, this was not the case for Autonomous Mobile Robots (AMRs), with AMR vendors extensively represented.
Generally, the event had the feel of being an innovation showcase, with talks and show stands exhibiting the future of robotics automation, rather than focusing on traditional form factors and use cases. The notorious Boston-based MassRobotics had a large presence at the event; opportunities existed for attendees to meet the accelerator’s successes and learn about their impressive innovations. MassRobotics also organized a job fair on the second day.
Below are some of the key trends and emerging areas I observed at the Robotics Summit & Expo 2024.
Medical Robotics
The event hosted many big players within the burgeoning medical robotics space. DeviceTalks, a MedTech expo hosted by WTWH media, took place at the same time as the Robotics event, with a lot of crossover between the two. MedTech exhibitors presented products spanning the length and breadth of the domain, including:
- Low-latency control for surgical Collaborative Robots (cobots)
- Autonomous robotics for performing ultrasound scans
- Patient digital twins for obtaining diagnostic information
- Medical product development and assistive Artificial Intelligence (AI)
- Tools for the software ecosystem to verify regulatory compliance
- Helping developers both curate compliant software and streamline updates
- Smart exoskeleton technology to aid with mobility and ease patient recovery
The market for medical robotics is large and growing; innovators within the space leverage the inherent safety of robotics form factors, insightful data streams, and improved stability and precision to provide medical practitioners with deeper insights into patient care and augment their capabilities as surgeons.
Robotics Components
Kollmorgen (and parent company Regal Rexnord) and Harmonic Drive were among the significant names in robotics component manufacturing present at the event. As highlighted in interviews with Elon Musk during the initial development cycle of Tesla’s humanoid robot, Optimus, components for exotic robot form factors cannot be bought off-the-shelf.
Robotics component manufacturers perform well at fabricating components in small volumes, but struggle at larger scales; this is simply because of the complexity of production and ecosystem support correlating to market size—there aren’t very many complex robots. Even now, ABI Research finds that a total of around 1 million robots are manufactured per year. Considering the diversity of form factors and vendors, there aren’t many orders from any single manufacturer, and no products are as complex as humanoids (there are 350 joints in the human body). This form factor, if it generates the boom in demand that proponents claim it will, could require the manufacture of hundreds of thousands of incredibly complex robots per year.
To meet such a demand, component manufacturers will need to scale and innovate dramatically. Competition is likely to grow between Hyundai (the new owner and commercial impetus behind Boston Dynamics) and Tesla. Both companies are car manufacturers—therefore possessing expertise in complex mass production and having immediate deployment sites. This makes these two innovators the ones to closely watch.
Figure 1: Boston Dynamics Spot Robot
(Source: ABI Research)
Vision & SLAM
Software vendors presented a host of solutions (generally Robot Operating System (ROS)-based) for localization and object classification. A standout was Indian startup CynLr’s demonstration that featured a large, two-armed cobot capable of picking and manipulating an object to find a barcode and place it in a correct bin. Picking is a significant use case for cobots and many vendors leverage Universal Robots’ collaborative arm to deliver value offerings to stakeholders. CynLr sells binocular vision systems and a host of Machine Learning (ML) classification and control algorithms. Therefore, CynLr falls into a similar category as SICK and Keyence, but with emerging robot form factors in mind (as opposed to classic industrial robots).
With the availability (and regular maintenance) of open-source Machine Vision (MV) repositories, companies need relatively small—although highly-skilled—software engineering teams to create vision, and even Simultaneous Location and Mapping (SLAM), solutions. Software vendors instead capitalize on providing solutions with advanced manipulation skills and reliable, cost-effective SLAM. Opteran, another exhibitor, is an example of a company that offers low-cost, reliable localization solutions for AMRs.
Autonomous Mobile Robots
The largest AMR vendors were present at the Robotics Summit, including Geek+, Locus Robotics, and Clearpath Robotics. Also present was Electric Sheep, a company that manufactures autonomous, two-wheeled, lawn mowing and leaf and litter clearing robots. Electric Sheep’s go-to-market is novel: it purchases landscaping companies and augments its services with the addition of their AMRs. Perhaps business models such as this are the key to robotics achieving market penetration and greater uptake beyond warehousing and material handling in manufacturing facilities—buy a company that operates in a certain domain and augment its operations with robots.
I spoke with Robust.AI—although it was not exhibiting at the event—about its recently unveiled AMR cart to aid milk-run-style picking. Robust.AI’s offering is of note because of the simple addition of a handle. The handle allows the AMR to be maneuvered by employees to a desired destination, Robust.AI’s Fleet Management Software (FMS) then updates the routing for the cart and permits it to resume autonomous operation. This simple innovation has allowed the company to provide truly collaborative mobile robots. A 2022 partnership with DHL catapulted Robust.AI into the public eye.
ROS and Data
Many robotics products are built using ROS due to its accessibility, simplicity, and compatibility with many data types/hardware drivers. The main communication method within an ROS network is the rosbag, a message package that is created at a source node (be it a hardware controller, sensor, general System-on-Chip (SoC), or web server) with a time stamp, data, and topics. A ROS node will then broadcast the bags, and another node will receive and unpack the data.
The prominence of rosbags has grown, as many device manufacturers now include the message type with their products (including Intel, NVIDIA, and robotics manufacturers such as KUKA and ABB), so being able to interpret and manipulate the messages is becoming increasingly important. An example of an innovator that uses ROS data is Roboto AI. This company offers a service to interpret rosbags (or other data formats), applying AI to look for key events, errors, or custom queries, providing insights in an easy and intuitive way.
Robotics Safety
Several vendors showcased products designed to address robotics safety. One example was 3Laws Robotics, a software company that creates geofences for robots of any form factor. This kind of physical safety will be key to letting robots loose on sites and will enable greater uptake. One can envision regulatory requirements stipulating geofencing as a requirement for autonomous drone use (despite the complexity of implementation when factoring in items like sensor drift and Global Positioning System (GPS) resolution).
Also present were BlackBerry QNX that showcased an extremely low-latency teleoperation cobot with haptic feedback (using a teleoperation product created by Haply Robotics, which was also at the event) and a geofence. Force feedback featured significantly across the cobot medical products on show—force and touch being critical sense capabilities that highlight the dichotomy between human surgeons and medical robots—factoring in tactile feedback is understandably a priority.
Figure 2 : Blackberry QNX Low-Latency Teleoperation of a Collaborative Robotic Arm
(Source: ABI Research)
To this end, I attended a talk delivered by Heba Khamis of Contactile, a company that specializes in force sensors for robot end effectors. Heba’s talk highlighted the importance of haptics for dexterous manipulation and control. Many of the exhibiters’ demo products featured Robotiq end effectors that also use intelligent force and grip detection.
Read More: Industrial Robot Safety Standards: A Short Rundown
Alternative Computing Methods for Robot Autonomy
Several software vendors at the Robotics Summit & Expo showcased ML systems for autonomous robots as alternatives to established paradigms and nascent generative AI. Standouts in this category include Opteran—a company that leverages biomimetic algorithms to achieve low-cost computational navigation (and decision-making) solutions for mobile robots—and ViVum, a company that uses neuromorphic microchips and algorithms to provide autonomy to drones, primarily for military applications.
Opteran recently achieved commercial success through a partnership with SAFELOG. These products come as a challenge to NVIDIA’s leadership within the robotics-compute space (likely not a serious challenge given the size and momentum behind the NVIDIA behemoth); hoping to gain favor by reducing associated computational costs, Bills of Materials (BOMs), and complexity. Given the current inflection point the robotics industry is experiencing and the few offered practical use cases for generative AI in robotics, alternative systems for machine intelligence have a fair shot at establishing themselves.
Figure 3: Sandra Skaff (NVIDIA), Juan Aparicio (Reshape Automation), Russ Tedrake (Toyota Research Institute), and Mike Oitzmam (WTWH Media) Discussing Generative AI's Impact on Robotics
(Source: ABI Research)
Robotics Simulation
Beyond the impressive hardware, robotics simulation technology also garnered a lot of attention. Notably, Dassault Systèmes’ medical division’s focus on patient simulation, Siemens’ talk on simulation and digital twins, and NVIDIA’s Gerard Andrew’s talk on the frontiers of simulation. Of note here was the scarcity of current digital twin deployments and the potential for symbiosis between product and factory in a virtual environment (analogously, Dassault demonstrated this with patient digital twins).
Figure 4: Electric Sheep's Verdie Robot Lawn Trimming
(Source: ABI Research)
Siemens highlighted the benefits of persistent digital twins with data, including the ease of commissioning new assets when digital infrastructure is already in place and actionable insights for reducing downtime, emissions, and wastage. NVIDIA continues to pursue the role of robotics champion; a double-barreled strategy sees NVIDIA achieving this goal through both simulation and edge compute for robots. NVIDIA’s Isaac Gym is a platform for training both machine vision algorithms and robot behaviors.
Many companies have trained their products in Isaac, including Electric Sheep, which used the platform to teach its flagship AMR to pick leaves and litter. Training robots in simulation is a niche but growing application. Innovatively, to maintain dominance of the simulation space, NVIDIA’s platform is very flexible. The platform allows for the integration of Application Programming Interfaces (APIs) of other key vendors in the robotics simulation space, including Visual Components and MathWorks, resulting in a customizable experience that will enable NVIDIA’s Isaac to adapt to many roboticists’ needs.
The incorporation of generative AI for robot control is also a main objective for NVIDIA. Isaac will enable generative AI for most robotic form factors: Isaac Manipulator will target cobots and classic industrial robots with the goal of teaching them how to interact with new objects. Meanwhile, Isaac Perceptor is used to train machine vision that can be deployed on any robot.
For more content about the latest robotics innovations, visit ABI Research’s Industrial, Collaborative & Commercial Robotics service page.
