MetaPusher: Meta Learning and Planning for Nonprehensile Manipulation of Unseen Objects with Rapid Online Adaption

Data-driven planar pushing methods have recently gained attention as they reduce manual engineering effort and improve generalization compared to analytical approaches. However, most prior work targets narrow capabilities (e.g., side switching, precision, or single-task training), limiting broader applicability. We present a model-based framework for non-prehensile tabletop pushing that uses a single learned model to address multiple tasks without retraining. Our approach employs a recurrent GRU-based architecture with additional non-linear layers to capture object-environment dynamics while ensuring stability. A tailored state-action representation enables the model to generalize across uncertain dynamics, variable push lengths, and diverse tasks. For control, we integrate the learned dynamics with a sampling-based Model Predictive Path Integral (MPPI) controller, which generates adaptive, task-oriented actions. This framework supports side switching, variable-length pushes, and objectives such as precise positioning, trajectory following, and obstacle avoidance. Training is performed in simulation with domain randomization to support sim-to-real transfer. We first evaluate the architecture through ablation studies, showing improved prediction accuracy and stable rollouts. We then validate the full system in simulation and real-world experiments using a Franka Panda robot with markerless tracking. Results demonstrate high success rates in precise positioning under strict thresholds and strong performance in trajectory tracking and obstacle avoidance. Moreover, multiple tasks are solved simply by changing the controller's objective function, without retraining. While our current focus is on a single object type, we extend the framework by training on wider push lengths and designing a balanced controller that reduces the number of steps for longer-horizon goals.

Non-prehensile manipulation is essential for handling thin, large, or otherwise ungraspable objects in unstructured settings. Prior planning and search-based methods often rely on ad-hoc manual designs or generate physically unrealizable motions by ignoring critical gripper properties, while training-based approaches are data-intensive and struggle to generalize to novel, out-of-distribution tasks. We propose a library-driven hierarchical planner (LDHP) that makes executability a first-class design goal: a top-tier contact-state planner proposes object-pose paths using MoveObject primitives, and a bottom-tier grasp planner synthesizes feasible grasp sequences with AdjustGrasp primitives; feasibility is certified by collision checks and quasi-static mechanics, and contact-sensitive segments are recovered via a bounded dichotomy refinement. This gripper-aware decomposition decouples object motion from grasp realizability, yields a task-agnostic pipeline that transfers across manipulation tasks and geometric variations without re-design, and exposes clean hooks for optional learned priors. Real-robot studies on zero-mobility lifting and slot insertion demonstrate consistent execution and robustness to shape and environment changes.

While pretrained robotic policies exhibit impressive capabilities in controlled environments, unobserved physical properties and dynamics require these policies to rapidly adapt during deployment. Existing test-time adaptation methods typically rely on sparse scalar rewards, failing to exploit the rich geometric and dynamic feedback from the environment during physical interaction. To address this challenge, we propose SCOUT, a dynamics-aware meta-learning framework that enables manipulation policies to rapidly adapt by continuously revising their internal beliefs about environment dynamics. Our approach couples an action-prediction policy with a forward dynamics model via a shared belief latent space. During meta-training, an inner loop updates this shared belief latent by minimizing the dynamics prediction error against the observed action outcome, while the outer loop optimizes the network for action selection. At deployment, this structure allows the agent to infer and adapt to unknown physical dynamics on the fly. By updating its latent belief based on action-outcome mismatches, the policy automatically adapts without risking catastrophic forgetting. We demonstrate that SCOUT significantly accelerates online adaptation across simulated manipulation benchmarks and achieves robust sim-to-real transfer in the real world. Project webiste can be found here: https://liy1shu.github.io/SCOUT/