Learning to Adapt: Representation-Based Reinforcement Learning for Multi-Task Skill Transfer

Reinforcement learning has achieved remarkable success in learning complex control policies, yet its applicability remains limited due to sample inefficiency and poor generalization across tasks. In this work, we propose RepMT-SAC, a framework for multi-task RL that enables efficient knowledge sharing and robust transfer to new tasks. RepMT-SAC uses spectral MDP decomposition to capture transferable dynamics, structuring the value function into a task-agnostic core with a minimal task-specific adjustment. This design allows for strong zero-shot performance on in-distribution tasks and rapid few-shot adaptation to out-of-distribution tasks. We evaluate RepMT-SAC on quadcopter trajectory-following tasks across in-distribution and out-of-distribution contexts, demonstrating that it outperforms baselines by up to 30%.

Language as a Sensor: Calibrated Spatial Belief Estimation in 3D Scenes from Natural Language

Robots deployed in human-centric environments routinely receive natural-language descriptions of spatial information ("I left my backpack on the table") that reference parts of the world beyond their perceptual field of view. Traditional metric-semantic mapping ignores this signal, while off-the-shelf multimodal models remain limited in 3D spatial reasoning and are not directly amenable to fusion with other sensor modalities. To convert language observations into a calibrated spatial distribution, we train a Language Sensor Model (LSM) that maps each utterance and its scene-graph context to a multimodal distribution, with mixture weights encoding referential ambiguity (e.g., "which table") and component covariances encoding spatial uncertainty (e.g., where "on the table" the target lies). We then introduce VL-Map (Vision-Language Metric-Semantic Mapping), a probabilistic framework that treats these language predictions as stochastic observations and fuses them with onboard perception within a unified belief map. On the VLA-3D benchmark as well as on a real-world mobile robot, LSM is the only language predictor whose covariance estimates remain within the calibrated regime; fused into VL-Map, it leads to more accurate predictions of the target object location (~70% more probability mass on the true target compared to the strongest foundation-model baseline).

Hardware-in-the-Loop for Characterization of Embedded State Estimation for Flying Microrobots

Autonomous flapping-wing micro-aerial vehicles (FWMAV) have a host of potential applications such as environmental monitoring, artificial pollination, and search and rescue operations. One of the challenges for achieving these applications is the implementation of an onboard sensor suite due to the small size and limited payload capacity of FWMAVs. The current solution for accurate state estimation is the use of offboard motion capture cameras, thus restricting vehicle operation to a special flight arena. In addition, the small payload capacity and highly non-linear oscillating dynamics of FWMAVs makes state estimation using onboard sensors challenging due to limited compute power and sensor noise. In this paper, we develop a novel hardware-in-the-loop (HWIL) testing pipeline that recreates flight trajectories of the Harvard RoboBee, a 100mg FWMAV. We apply this testing pipeline to evaluate a potential suite of sensors for robust altitude and attitude estimation by implementing and characterizing a Complimentary Extended Kalman Filter. The HWIL system includes a mechanical noise generator, such that both trajectories and oscillatinos can be emulated and evaluated. Our onboard sensing package works towards the future goal of enabling fully autonomous control for micro-aerial vehicles.