Multi-robot systems in cognitive factories: representation, reasoning, execution and monitoring

We propose the use of causality-based formal representation and automated reasoning methods from artificial intelligence to endow multiple teams of robots in a factory, with high-level cognitive capabilities, such as optimal planning and diagnostic reasoning. We present a framework that features bilateral interaction between task and motion planning, and embeds geometric reasoning in causal reasoning. We embed this planning framework inside an execution and monitoring framework and show its applicability on multi-robot systems. In particular, we focus on two domains that are relevant to cognitive factories: i) a manipulation domain with multiple robots working concurrently / co-operatively to achieve a common goal and ii) a factory domain with multiple teams of robots utilizing shared resources. In the manipulation domain two pantograph robots perform a complex task that requires true concurrency. The monitoring framework checks plan execution for two sorts of failures: collisions with unknown obstacles and change of the world due to human interventions. Depending on the cause of the failures, recovery is done by calling the motion planner (to find a different trajectory) or the causal reasoner (to find a new task plan). Therefore, recovery relies on not only motion planning but also causal reasoning. We extend our planning and monitoring framework for the factory domain with multiple teams of robots by introducing algorithms for finding optimal decoupled plans and diagnosing the cause of a failure/discrepancy (e.g., robots may get broken or tasks may get reassigned to teams). We show the applicability of these algorithms on an intelligent factory scenario through dynamic simulations and physical experiments.