Will AI robots take middle-class jobs?
A novel technique to improve the learning ability of robots that perform repetitive tasks
A plant-inspired controller that could facilitate the operation of robotic arms in real-world environments
Valmont Records Longest BVLOS Drone Flight on the Wings of T-Mobile 5G
#ICRA2023 awards finalists and winners
In this post we bring you all the paper awards finalists and winners presented during the 2023 edition of the IEEE International Conference on Robotics and Automation (ICRA). Congratulations to the winners and finalists!
ICRA 2023 Outstanding Paper
- Distributed Data-Driven Predictive Control for Multi-Agent Collaborative Legged Locomotion, by Fawcett, Randall; Amanzadeh, Leila; Kim, Jeeseop; Ames, Aaron; Akbari Hamed, Kaveh. (WINNER)
- Proficiency Self-Assessment without Breaking the Robot: Anomaly Detection using Assumption-Alignment Tracking from Safe Experiments, by Cao, Xuan; Crandall, Jacob W.; Pedersen, Ethan; Gautam, Alvika; Goodrich, Michael A.
- In-Hand Manipulation in Power Grasp: Design of an Adaptive Robot Hand with Active Surfaces, by Cai, Yilin; Yuan, Shenli.
ICRA 2023 Outstanding Automation Paper
- Towards Open-Set Material Recognition Using Robot Tactile Sensing, by Liu, Kun-Hong; Yang, Qianhui; Xie, Yu; Huang, Xiangyi.
- Target-Aware Implicit Mapping for Agricultural Crop Inspection, by Kelly, Shane; Riccardi, Alessandro; Marks, Elias Ariel; Magistri, Federico; Guadagnino, Tiziano; Chli, Margarita; Stachniss, Cyrill. (WINNER)
- Can Machines Garden? Systematically Comparing the AlphaGarden vs. Professional Horticulturalists, by Adebola, Simeon Oluwafunmilore; Parikh, Rishi; Presten, Mark; Sharma, Satvik; Aeron, Shrey; Rao, Ananth; Mukherjee, Sandeep; Qu, Tomson; Wistrom, Tina; Solowjow, Eugen; Goldberg, Ken.
ICRA 2023 Outstanding Student Paper
- M-EMBER: Tackling Long-Horizon Mobile Manipulation Via Factorized Domain Transfer, by Wu, Bohan; Martin-Martin, Roberto; Fei-Fei, Li.
- Robust Locomotion on Legged Robots through Planning on Motion Primitive Graphs, by Ubellacker, Wyatt; Ames, Aaron. (WINNER)
- Occlusion Reasoning for Skeleton Extraction of Self-Occluded Tree Canopies, by Kim, Chung Hee; Kantor, George.
ICRA 2023 Outstanding Deployed Systems Paper
- GUTS: Generalized Uncertainty-Aware Thompson Sampling for Multi-Agent Active Search, by Bakshi, Nikhil Angad; Gupta, Tejus; Ghods, Ramina; Schneider, Jeff. (WINNER)
- FRIDA: A Collaborative Robot Painter with a Differentiable, Real2Sim2Real Planning Environment, by Schaldenbrand, Peter; McCann, James; Oh, Jean.
ICRA 2023 Outstanding Dynamics and Control Paper
- Nonlinear Model Predictive Control of a 3D Hopping Robot: Leveraging Lie Group Integrators for Dynamically Stable Behaviors, by Csomay-Shanklin, Noel; Dorobantu, Victor; Ames, Aaron. (WINNER)
- Robust, High-Rate Trajectory Tracking on Insect-Scale Soft-Actuated Aerial Robots with Deep-Learned Tube MPC, by Tagliabue, Andrea; Hsiao, Yi-Hsuan; Fasel, Urban; Kutz, J. Nathan; Brunton, Steven; Chen, YuFeng; How, Jonathan.
- Autonomous Drifting with 3 Minutes of Data Via Learned Tire Models, by Djeumou, Franck; Goh, Jon; Topcu, Ufuk; Balachandran, Avinash.
ICRA 2023 Outstanding Healthcare and Medical Robotics Paper
- MRI-Powered Magnetic Miniature Capsule Robot with HIFU-Controlled On-Demand Drug Delivery, by Tiryaki, Mehmet Efe; Doğangün, Fatih; Dayan, Cem Balda; Wrede, Paul; Sitti, Metin.
- Real-Time Constrained 6D Object-Pose Tracking of an In-Hand Suture Needle for Minimally Invasive Robotic Surgery, by Chiu, Zih-Yun; Richter, Florian; Yip, Michael C. (WINNER)
- Exploring Robot-Assisted Optical Coherence Elastography for Surgical Palpation, by Chang, Yeonhee; Ahronovich, Elan; Simaan, Nabil; Song, Cheol.
ICRA 2023 Outstanding Locomotion Paper
- RAMP: Reaction-Aware Motion Planning of Multi-Legged Robots for Locomotion in Microgravity, by Ribeiro, Warley Francisco Rocha; Uno, Kentaro; Imai, Masazumi; Murase, Koki; Yoshida, Kazuya.
- Robust Locomotion on Legged Robots through Planning on Motion Primitive Graphs, by Ubellacker, Wyatt; Ames, Aaron.
- Multi-Segmented, Adaptive Feet for Versatile Legged Locomotion in Natural Terrain, by Chatterjee, Abhishek; Mo, An; Kiss, Bernadett; Gönen, Emre Cemal; Badri-Spröwitz, Alexander. (WINNER)
ICRA 2023 Outstanding Manipulation Paper
- M-EMBER: Tackling Long-Horizon Mobile Manipulation Via Factorized Domain Transfer, by Wu, Bohan; Martin-Martin, Roberto; Fei-Fei, Li.
- DexGraspNet: A Large-Scale Robotic Dexterous Grasp Dataset for General Objects Based on Simulation, by Wang, Ruicheng; Zhang, Jialiang; Chen, Jiayi; Xu, Yinzhen; Li, Puhao; Liu, Tengyu; Wang, He.
- In-Hand Manipulation in Power Grasp: Design of an Adaptive Robot Hand with Active Surfaces, by Cai, Yilin; Yuan, Shenli. (WINNER)
ICRA 2023 Outstanding Mechanisms and Design Paper
- New Bracket Polynomials Associated with the General Gough-Stewart Parallel Robot Singularities, by Thomas, Federico.
- A Compact, Two-Part Torsion Spring Architecture, by Bons, Zachary P; Thomas, Gray; Mooney, Luke; Rouse, Elliott.
- Contact Force Control with Continuously Compliant Robotic Legs, by Bendfeld, Robin; Remy, C. David. (WINNER)
ICRA 2023 Outstanding Multi-Robot Systems Paper
- Graph Neural Networks for Multi-Robot Active Information Acquisition, by Tzes, Mariliza; Bousias, Nikolaos; Chatzipantazis, Evangelos; Pappas, George J. (WINNER)
- Distributed Data-Driven Predictive Control for Multi-Agent Collaborative Legged Locomotion, by Fawcett, Randall; Amanzadeh, Leila; Kim, Jeeseop; Ames, Aaron; Akbari Hamed, Kaveh.
- GoRela: Go Relative for Viewpoint-Invariant Motion Forecasting, by Cui, Alexander; Casas Romero, Sergio; Wong, Kelvin; Suo, Simon; Urtasun, Raquel.
ICRA 2023 Outstanding Navigation Paper
- IMODE: Real-Time Incremental Monocular Dense Mapping Using Neural Field, by Matsuki, Hidenobu; Sucar, Edgar; Laidlow, Tristan; Wada, Kentaro; Scona, Raluca; Davison, Andrew J.
- SmartRainNet: Uncertainty Estimation for Laser Measurement in Rain, by Zhang, Chen; Huang, Zefan; Tung, Beatrix; Ang Jr, Marcelo H; Rus, Daniela. (WINNER)
- Online Whole-Body Motion Planning for Quadrotor Using Multi-Resolution Search, by Ren, Yunfan; Liang, Siqi; Zhu, Fangcheng; Lu, Guozheng; Zhang, Fu.
ICRA 2023 Outstanding Physical Human-Robot Interaction Paper
- A Control Approach for Human-Robot Ergonomic Payload Lifting, by Rapetti, Lorenzo; Sartore, Carlotta; Elobaid, Mohamed; Tirupachuri, Yeshasvi; Draicchio, Francesco; Kawakami, Tomohiro; Yoshiike, Takahide; Pucci, Daniele.
- Learning from Physical Human Feedback: An Object-Centric One-Shot Adaptation Method, by Shek, Alvin; Su, Bo Ying; Chn Rui; Liu, Changliu. (WINNER)
- Interactive Object Segmentation in 3D Point Clouds, by Kontogianni, Theodora; Celikkan, Ekin; Tang, Siyu; Schindler, Konrad.
ICRA 2023 Outstanding Planning Paper
- Obstacle-Aware Topological Planning Over Polyhedral Representation for Quadrotors, by Gao, Junjie; He, Fenghua; Zhang, Wei; Yao, Yu.
- Learning-Based Initialization of Trajectory Optimization for Path-Following Problems of Redundant Manipulators, by Yoon, Minsung; Kang, Mincheul; Park, Daehyung; Yoon, Sung-eui. (WINNER)
- A Multi-Step Dynamics Modeling Framework for Autonomous Driving in Multiple Environments, by Gibson, Jason; Vlahov, Bogdan; Fan, David D; Spieler, Patrick; Pastor, Daniel; Agha-mohammadi, Ali-akbar; Theodorou, Evangelos.
ICRA 2023 Outstanding Robot Learning Paper
- Code As Policies: Language Model Programs for Embodied Control, by Liang, Jacky; Huang, Wenlong; Xia, Fei; Xu, Peng; Hausman, Karol; Ichter, Brian; Florence, Peter; Zeng, Andy. (WINNER)
- Grounding Language with Visual Affordances Over Unstructured Data, by Mees, Oier; Borja Diaz, Jessica; Burgard, Wolfram.
- NeRF2Real: Sim2real Transfer of Vision-Guided Bipedal Motion Skills Using Neural Radiance Fields, by Byravan, Arunkumar; Humplik, Jan; Hasenclever, Leonard; Brussee, Arthur; Nori, Francesco; Haarnoja, Tuomas; Moran, Ben; Bohez, Steven; Sadeghi, Fereshteh; Vujatovic, Bojan; Heess, Nicolas.
ICRA 2023 Outstanding Sensors and Perception Paper
- Towards Consistent Batch State Estimation Using a Time-Correlated Measurement Noise Model, by Yoon, David Juny; Barfoot, Timothy.
- GMCR: Graph-Based Maximum Consensus Estimation for Point Cloud Registration, by Gentner, Michael; Murali, Prajval Kumar; Kaboli, Mohsen.
- Occlusion Reasoning for Skeleton Extraction of Self-Occluded Tree Canopies, by Kim, Chung Hee; Kantor, George. (WINNER)
Robot ‘chef’ learns to recreate recipes from watching food videos
Team makes electronic skin that can sense touch
Ranking the best humanoid robots of 2023
Is Rosie the Robot Maid from the Jetsons here yet? Several different types of humanoid are currently deployed commercially or in trials. We’ve come along way since the DARPA Robotics Challenge of 2015/2016, where the most popular footage was the blooper reels of robots falling over and failing to open doors or climb stairs.
The Avatar XPrize of 2019-2022 showcased some extremely sophisticated humanoids that certainly advanced the state of the art but the holy grail of humanoid robots is combining incredible sophistication into a sub $50,000 package. Why $50,000? Wouldn’t some companies pay a lot more? Then again, can’t we buy a car, also a very sophisticated device capable of partial autonomy that is 5 times the size of a humanoid, for less than $50,000? Why is this the benchmark for humanoids?
$50,000 is the annual wage for a single shift of labor at slightly more than $18/hour or minimum wage in every low wage industry. There is a terrible labor shortage and it is the dirty dull and dangerous jobs that are hardest for employers to fill. Companies that can afford to run two or more shifts a day also have more alternatives when it comes to filling their labor gaps. It’s the small to medium size enterprise that is suffering the most in our current economic and demographic conditions.
We don’t need a Six Million Dollar Man.
We need a $50,000 humanoid.
The roll out of sophisticated new robots and how we integrate them into society is at the heart of my early research and my current roles as the Managing Director of Silicon Valley Robotics (explain), VP of Global Robotics for AMT (explain) and the VP of Industrial Activities for the IEEE Robotics and Automation Society (explain).
As more and more companies announce their work towards the affordable humanoid robot, I wanted to create a reference chart for myself, and realized that it might be of interest to others as well. The ranking system is just my own opinion and it will be fascinating to see who succeeds and progresses over the next few years. Enjoy this overview and make up your own minds as to which humanoid robot is really the best.
Who’s in the running? (in alphabetical order by company not robot)
- 1x – Eve
- Aeolus Robotics – Eva
- Agility Robotics – Digit
- Apptronik – Astra
- Boston Dynamics – Atlas
- Comma.ai – body
- Devanthro – Robody
- Engineered Arts – Ameca
- Figure – Figure01
- Giant.ai – Universal Worker
- IIT – ErgoCub
- PAL – Reem-C
- Prosper Robotics – Alfie
- Sanctuary – Phoenix
- Tesla – Optimus
- Toyota – T-HR3
Who isn’t in the running?
Hollywood Humanoids are one off robots for the purpose of entertainment, like Sophia from Hanson Robotics, Xoxe from AI Life, or Beonmi from Beyond Imagination. ….
It’s too hard for me to validate that they exist, work as advertized, and what the specifications are.
Love them but they have a different purpose. Only robots with commercial deployment plans, and ideally, a price tag and a date in 2023 or 2024 when they’ll be available for purchase, if they aren’t already being sold.
I also love robots that work like a humanoid but don’t look human-like. We saw some examples in the DARPA Robotics Challenge, most notably RoboSimian. Once we go down that route, all quadrupeds, and multi-armed robots or wheeled humanlike robots, would qualify. Who knew there were so many robots!
Who have I missed?
I’m hoping to crowdsource some more great robots :)
Read the original article on Substack.
Fronius Perfect Welding
Serve Unveils Commercial Deal with Uber to Enable Scaling of Robotic Delivery
Robot Talk Episode 51 – James Kell
Claire chatted to James Kell from Jacobs Engineering UK all about civil infrastructure, nuclear robotics and jet engine inspection.
James Kell is Robotics Technical Director at Jacobs Engineering UK. He is responsible for the internal robotics technical strategy and is actively looking to apply the decades worth of robotics and remote handling experience that Jacobs has to the wider market in other sectors. The overall intention is to extend the life of critical national infrastructure like roads, rail, and water. James’s previously worked for Rolls-Royce where his role was to develop keyhole surgery technologies to service jet engines. James is also a member of the Robotics Growth Partnership, supported by BEIS.
Barkour benchmark measures quadruped robot agility
Welcome to the Age of the Cobot
Automate 2023 recap and the receding horizon problem
“Thirty million developers” are the answer to driving billion-dollar robot startups, exclaimed Eliot Horowitz of Viam last week at Automate. The hushed crowd of about 200 hardware entrepreneurs listened intensely to MongoDB‘s founder and former CTO (a $20Bn success story). Now, Horowitz aims to take the same approach that he took to democratizing cloud data applications to mechatronics. As I nudged him with questions about how his new platform will speed complex robot deployments to market, he shared his vision of the Viam developer army (currently 1,000+ strong) creating applications that can be seamlessly downloaded on the fly to any system and workflow. Unlike RoS which is primarily targeted to the current community of roboticists, Viam is luring the engineers that birthed ChatGPT to revolutionize uncrewed systems with new mechanical tasks addressing everyday needs. Imagine generative AI prompts for SLAM, gripping, computer vision, and other highly manipulative tasks with drag-and-drop ease.
Interviewing Horowitz recalled my discussion a few months back with Dr. Hal Thorsrud of Anges Scott College in Georgia. Professor Thorsrud teaches a novel philosophy course at this Liberal Arts institution on the “Introduction to Artificial Intelligence.” Similar to Horowitz, Thorsrud envisions an automated world whereby his graduates would be critical in thinking through the ethical applications of robots and AI. “Ethics has to become an engineering problem which is fascinating because, I mean, the idea is that we need to figure out how we can actually encode our ethical values into these systems. So they will abide by our values in order to pursue what we deemed to be good,” remarked Thorsrud.
According to Thorsrud’s syllabus, the class begins: “with a brief survey of positions in the philosophy of mind in order to better understand the concept of intelligence and to formulate the default position of most AI research, namely Computationalism. We then examine questions such as ‘What is a computer?’, ‘What makes a function or number computable?’, ‘What are algorithms and how do they differ from heuristics?’ We will consider fundamental issues in AI such as what makes a system intelligent, and whether computers can have minds. Finally, we will explore some of the ethical challenges that face AI such as whether intelligent artificial systems should be relied upon to make important decisions that affect our lives, and whether we should create such systems in the first place.”
In explaining the origins of his course, Dr. Thorsrud recalled, “A lot of my students are already interested in philosophy. They just don’t know it. And so, in fact, just recently my department has joined forces with neuroscience. We’re no longer a Philosophy Department. We’re now the Philosophy of Neuroscience, and now the Department of Law, Neuroscience, and Philosophy. Because these students are interested in mind, they are interested in intelligence, but they don’t realize that philosophy has been dealing with an attempt to understand the nature of mind from the very beginning, and the nature of intelligence from the very beginning. So we have a lot to offer these students this question of how to reach them. So that’s what kind of started me off on this different path, and, in the meantime think the same is true of artificial intelligence.”
Thorsrud elaborated that this introductory course is only the first step in a wider AI curriculum at Anges Scott as the confluence between endeavors like Viam and ChatGPT collide in the coming years to move the automation industry at hyperspeed. Already, the AI Philosopher sees how GPT is challenging humans to stand out, “The massive growth in the training data and the parameters, the weights that were that the machine learning was operating on really paid off.” He continued to illustrate how dystopian fears are unfounded, “I mean, we have a tendency to anthropomorphize things like ChatGPT and it’s understandable. But as far as I can tell, it’s, it’s a long way from the intelligence of my dog, a long, long way.” He is realistic about the speed of adoption, “Well, as a philosopher, they’re never going to be able to get to the point where they can give me a credible adjudication, because human judgment you know it is. And, this is another example of the ever-present receding horizon problem. First, you know that computers will never be able to beat a human at chess. Okay, fine computers will never be able to beat a human at GO. Fine computers will never be able to write. And so we keep setting these limits down, and then surpassing them.”
At Automate, I had the chance to catch up with ff Venture Capital portfolio company, PlusOne Robotics, and its amazing founder, Erik Nieves. While the talk in the theater was about the future, Nieves illustrated on the floor what is happening today. Impressively the startup is close to one million picks of depalletizing packages and sorting goods for the likes of FedEx and other leading providers of shipping & logistics. PlusOne’s proprietary computer vision co-bot platform is not waiting for the next generation of developers to join the ranks, but building its own intelligent protocols to increase efficiencies on the front lines of e-commerce fulfillment.
As Brian Marflak, of FedEx, remarked, “The technology in these depalletizing arms helps us move certain shipments that would otherwise take up valuable resources to manually offload. Having these systems installed allows team members to perform more skilled tasks such as loading and unloading airplanes and trucks. This has been a great opportunity for robotics to complement our existing team members and help them complete tasks more efficiently.”
Markflak’s sentiment was shared by the 25,000+ attendees of Automate that filled the entire Detroit Convention Center. A big backdrop of the show was how macro labor trends and shortages are exasperating the push towards automation (and thus moving the horizon even further). According to the most recent reports, close to 20% of all US retail sales are driven online, with over 20 billion packages being shipped every year growing at an annual rate of 25%. This means even if the e-commerce industry is able to hire a million more workers, there are not enough (organic) hands to keep up. As Nieves puts it, “The growth of e-commerce has placed tremendous pressure on shipping responsiveness and scalability that has significantly exacerbated labor and capacity issues. Automation is key, but keeping a human-in-the-loop is essential to running a business 24/7 with greater speed and fewer errors. With the ongoing labor shortages, I believe we’ll see an increase in the adoption of Robots-as-a-Service (RaaS) to lower capital expenditures and deploy automation on a subscription basis.” Get ready for Automate 2024, as the convention moves for the first time to an annual gathering!