CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Tasks

General-purpose robots coexisting with humans in their environment must learn to relate human language to their perceptions and actions to be useful in a range of daily tasks. Moreover, they need to acquire a diverse repertoire of general-purpose skills that allow composing long-horizon tasks by following unconstrained language instructions. In this letter, we present Composing Actions from Language and Vision (CALVIN) (Composing Actions from Language and Vision), an open-source simulated benchmark to learn long-horizon language-conditioned tasks. Our aim is to make it possible to develop agents that can solve many robotic manipulation tasks over a long horizon, from onboard sensors, and specified only via human language. CALVIN tasks are more complex in terms of sequence length, action space, and language than existing vision-and-language task datasets and supports flexible specification of sensor suites. We evaluate the agents in zero-shot to novel language instructions and to novel environments. We show that a baseline model based on multi-context imitation learning performs poorly on CALVIN, suggesting that there is significant room for developing innovative agents that learn to relate human language to their world models with this benchmark.

Automated summary

This article presents an open-source simulated benchmark called CALVIN, which is used to learn long-horizon tasks. CALVIN tasks are more complex and support flexible specification of sensor suites. Evaluation of the agents shows that a baseline model based on multi-context imitation learning performs poorly on CALVIN.