The Universal Robots RTDE communication interface is well-known in literature and it was used in several works. In [5] and [6] RTDE was adopted to control UR cobots. In [7], [8], and [9], the RTDE interface was used only for data acquisition. To facilitate the development of external applications for UR cobots, various higher-level software interfaces and drivers have been proposed based on RTDE. In addition to the official software interface by Universal Robots (ur_client_li-brary), a few alternatives have been developed by third-parties. One of these software interfaces is ur_rtde [4] by SDU Robotics, which was used in this work. Another similar interface is python-urx [10], which is a Python interface for tasks that do not require high control frequency.

In this article, we focus on the pre-training of visual autonomous driving agents in the context of imitation learning. Current methods often rely on a classification-based pre-training, which we hypothesise to be holding back from extending capabilities of implicit image understanding. We propose pre-training the visual encoder of a driving agent using the self-distillation with no labels (DINO) method, which relies on a self-supervised learning paradigm.% and is trained on an unrelated task. Our experiments in CARLA environment in accordance with the Leaderboard benchmark reveal that the proposed pre-training is more efficient than classification-based pre-training, and is on par with the recently proposed pre-training based on visual place recognition (VPRPre).

In robotics, the use of Large Language Models (LLMs) is becoming prevalent, especially for understanding human commands. In particular, LLMs are utilized as domain-agnostic task planners for high-level human commands. LLMs are capable of Chain-of-Thought (CoT) reasoning, and this allows LLMs to be task planners. However, we need to consider that modern robots still struggle to perform complex actions, and the domains where robots can be deployed are limited in practice. This leads us to pose a question: If small LMs can be trained to reason in chains within a single domain, would even small LMs be good task planners for the robots? To train smaller LMs to reason in chains, we build `COmmand-STeps datasets' (COST) consisting of high-level commands along with corresponding actionable low-level steps, via LLMs. We release not only our datasets but also the prompt templates used to generate them, to allow anyone to build datasets for their domain. We compare GPT3.5 and GPT4 with the finetuned GPT2 for task domains, in tabletop and kitchen environments, and the result shows that GPT2-medium is comparable to GPT3.5 for task planning in a specific domain. Our dataset, code, and more output samples can be found in https://github.com/Gawon-Choi/small-LMs-Task-Planning

In this paper, we present a system to train driving policies from experiences collected not just from the ego-vehicle, but all vehicles that it observes. This system uses the behaviors of other agents to create more diverse driving scenarios without collecting additional data. The main difficulty in learning from other vehicles is that there is no sensor information. We use a set of supervisory tasks to learn an intermediate representation that is invariant to the viewpoint of the controlling vehicle. This not only provides a richer signal at training time but also allows more complex reasoning during inference. Learning how all vehicles drive helps predict their behavior at test time and can avoid collisions. We evaluate this system in closed-loop driving simulations. Our system outperforms all prior methods on the public CARLA Leaderboard by a wide margin, improving driving score by 25 and route completion rate by 24 points. Our method won the 2021 CARLA Autonomous Driving challenge. Code and data are available at https://github.com/dotchen/LAV.

The CEO of Boston Dynamics says more warehouse operators are considering a robot workforce after COVID-19 exposed health vulnerabilities at logistics hubs. His comments come as Amazon (AMZN) warns it could run out of workers by 2024. “They have almost 100 per cent turn-over in logistics jobs like picking and packing boxes,” Robert Playter told Yahoo Finance Canada at the Collision tech conference in Toronto. “We’ve definitely seen [with] our industrial or warehouse customers [that] interest in robotics has only increased during the pandemic.” Boston Dynamics has shown its “Stretch” robot is smart enough to react to a stack of boxes suddenly falling over, and clean up the mess. The company plans to release a new robot every three-to-five years aimed at mastering a new workplace task. But Playter says the key is Boston Dynamics looks for the sweet spot between what the labour market needs, and what its robots are capable of doing. “The next robot, which we hope will come out in a few years, will probably be pushing in the direction of more dexterous manipulation tasks, perhaps in a manufacturing environment,” he said. Late last year, the Hyundai Motor Company (HYMTF) acquired an 80 per cent stake. Playter says the new majority owner will help commercialize its robots with its expertise in large-scale manufacturing. “They're going to help us create these things more efficiently, and lower the cost,'' he said. “By the end of this year, we'll have about 1,000 robots out with customers. So we're seeing strong interest.” Asked if robots will push human labour out of warehouses, he said, “I think a lot of the manual work will be done by robots. But robots aren't as smart as people yet, and you have to deal with unexpected circumstances.” Playter says he envisions an “up-skilling path” for workers to become robot operators. “The robot, its intelligence, handles a lot of the complexities. You just give it very high-level commands about what to do, sort of point it in a direction, or lay down a route. And it will autonomously do that work on its own,” he said. “It won’t take a college degree to operate them.” Jeff Lagerquist is a senior reporter at Yahoo Finance Canada. Follow him on Twitter @jefflagerquist. Download the Yahoo Finance app, available for Apple and Android.

While Reinforcement Learning (RL) approaches lead to significant achievements in a variety of areas in recent history, natural language tasks remained mostly unaffected, due to the compositional and combinatorial nature that makes them notoriously hard to optimize. With the emerging field of Text-Based Games (TBGs), researchers try to bridge this gap. Inspired by the success of RL algorithms on Atari games, the idea is to develop new methods in a restricted game world and then gradually move to more complex environments. Previous work in the area of TBGs has mainly focused on solving individual games. We, however, consider the task of designing an agent that not just succeeds in a single game, but performs well across a whole family of games, sharing the same theme. In this work, we present our deep RL agent--LeDeepChef--that shows generalization capabilities to never-before-seen games of the same family with different environments and task descriptions. The agent participated in Microsoft Research's "First TextWorld Problems: A Language and Reinforcement Learning Challenge" and outperformed all but one competitor on the final test set. The games from the challenge all share the same theme, namely cooking in a modern house environment, but differ significantly in the arrangement of the rooms, the presented objects, and the specific goal (recipe to cook). To build an agent that achieves high scores across a whole family of games, we use an actor-critic framework and prune the action-space by using ideas from hierarchical reinforcement learning and a specialized module trained on a recipe database.

Brain-computer interfaces (BCIs) allow a user to directly control devices such as cursors and robots using brain signals. Non-invasive BCIs, e.g., those based on electroencephalographic (EEG) signals recorded from the scalp, suffer from low signal-to-noise ratio which limits the bandwidth of control. Invasive BCIs allow fine-grained control but can leave users exhausted since control is typically exerted on a moment-by-moment basis. In this paper, we address these problems by proposing a new adaptive hierarchical architecture for brain-computer interfacing. The approach allows a user to teach the BCI new skills on-the-fly; these learned skills are later invoked directly as high-level commands, relieving the user of tedious low-level control. We report results from four subjects who used a hierarchical EEG-based BCI to successfully train and control a humanoid robot in a virtual home environment. Gaussian processes were used for learning high-level commands, allowing a BCI to switch between autonomous and user-guided modes based on the current estimate of uncertainty. We also report the first instance of multi-tasking in a BCI, involving simultaneous control of two different devices by a single user. Our results suggest that hierarchical BCIs can provide a flexible and robust way of controlling complex robotic devices in real-world environments.