We present Learning to Drive (L2D), a low-cost benchmark for real-world reinforcement learning (RL). L2D involves a simple and reproducible experimental setup where an RL agent has to learn to drive a Donkey car around three miniature tracks, given only monocular image observations and speed of the car. The agent has to learn to drive from disengagements, which occurs when it drives off the track. We present and open-source our training pipeline, which makes it straightforward to apply any existing RL algorithm to the task of autonomous driving with a Donkey car. We test imitation learning, state-of-the-art model-free, and model-based algorithms on the proposed L2D benchmark. Our results show that existing RL algorithms can learn to drive the car from scratch in less than five minutes of interaction. We demonstrate that RL algorithms can learn from sparse and noisy disengagement to drive even faster than imitation learning and a human operator.
The rapid development of Industry 4.0 expands industrial robots from isolated machines to collaborative agents. Recognizing this paradigm shift, design educators are increasingly aware of the need to prepare students for effectively engaging with emerging robotic technologies and shaping future human-robot collaboration (HRC) projects. However, systematic and structured resources to guide these student designers in approaching, conceptualizing, prototyping and implementing HRC remain scarce. This article introduces a novel HRC design toolkit tailored for student designers that incorporates industrial constraints, role assignment, and guided communication into a step-by-step design process. Evaluations with 13 student interaction designers showed that the proposed toolkit clarified task flows, facilitated more intuitive robot behaviors, and bridged the gap between theoretical knowledge and practical implementations. Based on participants’ feedback, we improved the toolkit with simpler visual aids and improved guidance, further lowering the threshold for creating collaborative tasks. Finally, we discussed the applications and contributions of this revised design toolkit to industrial robotics education.
Close-proximity human-robot collaboration (HRC) refers to scenarios where humans and robots work together for the same aim at the same time and space without safety fence. Few design tools currently exist to assist designers to think through how to design human-robot interactions, with a focus on turn-taking, coordination, and flow of intention. We present a design toolkit for designers to explore, specify, analyse and document the unfolding interaction between human(s) and robot(s) through a model of interaction where the roles of leader and follower shape the forms of collaborative behaviour. We illustrate our design toolkit in action and demonstrate how it can be used to analyse existing cases of HRC, or aid in exploring new applications and scenarios. We exemplify its use for the case of collaborative drawing, where human and robot jointly create a single drawing on a shared canvas in close proximity.