Learning Action Duration and Synergy in Task Planning for Human-Robot Collaboration

Abstract

A good estimation of the actions' cost is key in task planning for human-robot collaboration. The duration of an action depends on agents' capabilities and the correlation between actions performed simultaneously by the human and the robot. This paper proposes an approach to learning actions' costs and coupling between actions executed concurrently by humans and robots. We leverage the information from past executions to learn the average duration of each action and a synergy coefficient representing the effect of an action performed by the human on the duration of the action performed by the robot (and vice versa). We implement the proposed method in a simulated scenario where both agents can access the same area simultaneously. Safety measures require the robot to slow down when the human is close, denoting a bad synergy of tasks operating in the same area. We show that our approach can learn such bad couplings so that a task planner can leverage this information to find better plans.

Optimal Task and Motion Planning and Execution for Multiagent Systems in Dynamic Environments

“…Future works will focus on integrating learning techniques to refine the process model through experience and speed up the search for optimal plans [48]. For example, [49] presents a preliminary study on learning task duration and human-robot synergy via linear regression. Further investigation will also address the use of other optimization objectives (e.g., taking into account human factors and user preferences [50] or agent-agent communication [51]).…”

Learning and planning for optimal synergistic human-robot coordination in manufacturing contexts

“…The concept of synergy for each pair of human-robot tasks plays a crucial role in this article and must be estimated from experience. [8] proposed a linear model that relates the measured duration of human tasks to the level of parallelism of each task the robot performs through the synergy terms defined in (1). Specifically, for a batch of n realizations of task τ k executed by the human (H), we can write:…”

“…Instead of learning the synergy via linear regression as in [8], we estimate the synergy coefficients using a Bayesian approach.…”

“…The software architecture resembles that proposed in [8] and is shown in Figure 1. At the highest level, a task planning module solves the allocation and scheduling problem.…”

“…• We propose a method for estimating synergy between pairs of human-robot tasks based on knowledge gained from past process executions. Synergy terms encode the slowdown to the average execution duration of a robot task owed to the parallelism of a task performed by the human operator [8]. We use Bayesian regression for coefficient estimation to exploit a priori knowledge of the effect of the agents' tasks on each other.…”