We consider the problem of sequential manipulation and tool-use planning in domains that include physical interactions such as hitting and throwing. The approach integrates a Task And Motion Planning formulation with primitives that either impose stable kinematic constraints or differentiable dynamical and impulse exchange constraints at the path optimization level. We demonstrate our approach on a variety of physical puzzles that involve tool use and dynamic interactions. We then compare manipulation sequences generated by our approach to human actions on analogous tasks, suggesting future directions and illuminating current limitations.

Many animals, and an increasing number of artificial agents, display sophisticated capabilities to perceive and manipulate objects. But human beings remain distinctive in their capacity for flexible, creative tool use—using objects in new ways to act on the world, achieve a goal, or solve a problem. To study this type of general physical problem solving, we introduce the Virtual Tools game. In this game, people solve a large range of challenging physical puzzles in just a handful of attempts. We propose that the flexibility of human physical problem solving rests on an ability to imagine the effects of hypothesized actions, while the efficiency of human search arises from rich action priors which are updated via observations of the world. We instantiate these components in the “sample, simulate, update” (SSUP) model and show that it captures human performance across 30 levels of the Virtual Tools game. More broadly, this model provides a mechanism for explaining how people condense general physical knowledge into actionable, task-specific plans to achieve flexible and efficient physical problem solving.

We present Residual Policy Learning (RPL): a simple method for improving nondifferentiable policies using model-free deep reinforcement learning. RPL thrives in complex robotic manipulation tasks where good but imperfect controllers are available. In these tasks, reinforcement learning from scratch remains data-inefficient or intractable, but learning a residual on top of the initial controller can yield substantial improvements. We study RPL in six challenging MuJoCo tasks involving partial observability, sensor noise, model misspecification, and controller miscalibration. For initial controllers, we consider both handdesigned policies and model-predictive controllers with known or learned transition models. By combining learning with control algorithms, RPL can perform long-horizon, sparse-reward tasks for which reinforcement learning alone fails. Moreover, we find that RPL consistently and substantially improves on the initial controllers. We argue that RPL is a promising approach for combining the complementary strengths of deep reinforcement learning and robotic control, pushing the boundaries of what either can achieve independently 1 . * Equal contribution.1 Video and code at https://k-r-allen.github.io/residual-policy-learning/.

Humans can learn many novel tasks from a very small number (1–5) of demonstrations, in stark contrast to the data requirements of nearly tabula rasa deep learning methods. We propose an expressive class of policies, a strong but general prior, and a learning algorithm that, together, can learn interesting policies from very few examples. We represent policies as logical combinations of programs drawn from a domain-specific language (DSL), define a prior over policies with a probabilistic grammar, and derive an approximate Bayesian inference algorithm to learn policies from demonstrations. In experiments, we study six strategy games played on a 2D grid with one shared DSL. After a few demonstrations of each game, the inferred policies generalize to new game instances that differ substantially from the demonstrations. Our policy learning is 20–1,000x more data efficient than convolutional and fully convolutional policy learning and many orders of magnitude more computationally efficient than vanilla program induction. We argue that the proposed method is an apt choice for tasks that have scarce training data and feature significant, structured variation between task instances.

Social insect colonies use interactions among workers to regulate collective behavior. Harvester ant foragers interact in a chamber just inside the nest entrance, here called the 'entrance chamber'. Previous studies of the activation of foragers in red harvester ants show that an outgoing forager inside the nest experiences an increase in brief antennal contacts before it leaves the nest to forage. Here we compare the interaction rate experienced by foragers that left the nest and ants that did not. We found that ants in the entrance chamber that leave the nest to forage experienced more interactions than ants that descend to the deeper nest without foraging. Additionally, we found that the availability of foragers in the entrance chamber is associated with the rate of forager return. An increase in the rate of forager return leads to an increase in the rate at which ants descend to the deeper nest, which then stimulates more ants to ascend into the entrance chamber. Thus a higher rate of forager return leads to more available foragers in the entrance chamber. The highest density of interactions occurs near the nest entrance and the entrances of the tunnels from the entrance chamber to the deeper nest. Local interactions with returning foragers regulate both the activation of waiting foragers and the number of foragers available to be activated.