Revisiting Reward Design and Evaluation for Robust Humanoid Standing and Walking

A necessary capability for humanoid robots is the ability to stand and walk while rejecting natural disturbances. Recent progress has been made using sim-to-real reinforcement learning (RL) to train such locomotion controllers, with approaches differing mainly in their reward functions. However, prior works lack a clear method to systematically test new reward functions and compare controller performance through repeatable experiments. This limits our understanding of the trade-offs between approaches and hinders progress. To address this, we propose a low-cost, quantitative benchmarking method to evaluate and compare the real-world performance of standing and walking (SaW) controllers on metrics like command following, disturbance recovery, and energy efficiency. We also revisit reward function design and construct a minimally constraining reward function to train SaW controllers. We experimentally verify that our benchmarking framework can identify areas for improvement, which can be systematically addressed to enhance the policies. We also compare our new controller to state-of-the-art controllers on the Digit humanoid robot. The results provide clear quantitative trade-offs among the controllers and suggest directions for future improvements to the reward functions and expansion of the benchmarks.

Attention-based Models for Snow-Water Equivalent Prediction

Snow Water-Equivalent (SWE) -- the amount of water available if snowpack is melted -- is a key decision variable used by water management agencies to make irrigation, flood control, power generation and drought management decisions. SWE values vary spatiotemporally -- affected by weather, topography and other environmental factors. While daily SWE can be measured by Snow Telemetry (SNOTEL) stations with requisite instrumentation, such stations are spatially sparse requiring interpolation techniques to create spatiotemporally complete data. While recent efforts have explored machine learning (ML) for SWE prediction, a number of recent ML advances have yet to be considered. The main contribution of this paper is to explore one such ML advance, attention mechanisms, for SWE prediction. Our hypothesis is that attention has a unique ability to capture and exploit correlations that may exist across locations or the temporal spectrum (or both). We present a generic attention-based modeling framework for SWE prediction and adapt it to capture spatial attention and temporal attention. Our experimental results on 323 SNOTEL stations in the Western U.S. demonstrate that our attention-based models outperform other machine learning approaches. We also provide key results highlighting the differences between spatial and temporal attention in this context and a roadmap toward deployment for generating spatially-complete SWE maps.

Sim-to-Real Learning for Humanoid Box Loco-Manipulation

In this work we propose a learning-based approach to box loco-manipulation for a humanoid robot. This is a particularly challenging problem due to the need for whole-body coordination in order to lift boxes of varying weight, position, and orientation while maintaining balance. To address this challenge, we present a sim-to-real reinforcement learning approach for training general box pickup and carrying skills for the bipedal robot Digit. Our reward functions are designed to produce the desired interactions with the box while also valuing balance and gait quality. We combine the learned skills into a full system for box loco-manipulation to achieve the task of moving boxes from one table to another with a variety of sizes, weights, and initial configurations. In addition to quantitative simulation results, we demonstrate successful sim-to-real transfer on the humanoid r

Learning Vision-Based Bipedal Locomotion for Challenging Terrain

Reinforcement learning (RL) for bipedal locomotion has recently demonstrated robust gaits over moderate terrains using only proprioceptive sensing. However, such blind controllers will fail in environments where robots must anticipate and adapt to local terrain, which requires visual perception. In this paper, we propose a fully-learned system that allows bipedal robots to react to local terrain while maintaining commanded travel speed and direction. Our approach first trains a controller in simulation using a heightmap expressed in the robot's local frame. Next, data is collected in simulation to train a heightmap predictor, whose input is the history of depth images and robot states. We demonstrate that with appropriate domain randomization, this approach allows for successful sim-to-real transfer with no explicit pose estimation and no fine-tuning using real-world data. To the best of our knowledge, this is the first example of sim-to-real learning for vision-based bipedal locomotion over challenging terrains.

Multi-Task Learning for Budbreak Prediction

Grapevine budbreak is a key phenological stage of seasonal development, which serves as a signal for the onset of active growth. This is also when grape plants are most vulnerable to damage from freezing temperatures. Hence, it is important for winegrowers to anticipate the day of budbreak occurrence to protect their vineyards from late spring frost events. This work investigates deep learning for budbreak prediction using data collected for multiple grape cultivars. While some cultivars have over 30 seasons of data others have as little as 4 seasons, which can adversely impact prediction accuracy. To address this issue, we investigate multi-task learning, which combines data across all cultivars to make predictions for individual cultivars. Our main result shows that several variants of multi-task learning are all able to significantly improve prediction accuracy compared to learning for each cultivar independently.

Grape Cold Hardiness Prediction via Multi-Task Learning

Cold temperatures during fall and spring have the potential to cause frost damage to grapevines and other fruit plants, which can significantly decrease harvest yields. To help prevent these losses, farmers deploy expensive frost mitigation measures, such as, sprinklers, heaters, and wind machines, when they judge that damage may occur. This judgment, however, is challenging because the cold hardiness of plants changes throughout the dormancy period and it is difficult to directly measure. This has led scientists to develop cold hardiness prediction models that can be tuned to different grape cultivars based on laborious field measurement data. In this paper, we study whether deep-learning models can improve cold hardiness prediction for grapes based on data that has been collected over a 30-year time period. A key challenge is that the amount of data per cultivar is highly variable, with some cultivars having only a small amount. For this purpose, we investigate the use of multi-task learning to leverage data across cultivars in order to improve prediction performance for individual cultivars. We evaluate a number of multi-task learning approaches and show that the highest performing approach is able to significantly improve over learning for single cultivars and outperforms the current state-of-the-art scientific model for most cultivars.

Oracle Analysis of Representations for Deep Open Set Detection

The problem of detecting a novel class at run time is known as Open Set Detection & is important for various real-world applications like medical application, autonomous driving, etc. Open Set Detection within context of deep learning involves solving two problems: (i) Must map the input images into a latent representation that contains enough information to detect the outliers, and (ii) Must learn an anomaly scoring function that can extract this information from the latent representation to identify the anomalies. Research in deep anomaly detection methods has progressed slowly. One reason may be that most papers simultaneously introduce new representation learning techniques and new anomaly scoring approaches. The goal of this work is to improve this methodology by providing ways of separately measuring the effectiveness of the representation learning and anomaly scoring. This work makes two methodological contributions. The first is to introduce the notion of Oracle anomaly detection for quantifying the information available in a learned latent representation. The second is to introduce Oracle representation learning, which produces a representation that is guaranteed to be sufficient for accurate anomaly detection. These two techniques help researchers to separate the quality of the learned representation from the performance of the anomaly scoring mechanism so that they can debug and improve their systems. The methods also provide an upper limit on how much open category detection can be improved through better anomaly scoring mechanisms. The combination of the two oracles gives an upper limit on the performance that any open category detection method could achieve. This work introduces these two oracle techniques and demonstrates their utility by applying them to several leading open category detection methods.

Dynamic Bipedal Maneuvers through Sim-to-Real Reinforcement Learning

For legged robots to match the athletic capabilities of humans and animals, they must not only produce robust periodic walking and running, but also seamlessly switch between nominal locomotion gaits and more specialized transient maneuvers. Despite recent advancements in controls of bipedal robots, there has been little focus on producing highly dynamic behaviors. Recent work utilizing reinforcement learning to produce policies for control of legged robots have demonstrated success in producing robust walking behaviors. However, these learned policies have difficulty expressing a multitude of different behaviors on a single network. Inspired by conventional optimization-based control techniques for legged robots, this work applies a recurrent policy to execute four-step, 90 degree turns trained using reference data generated from optimized single rigid body model trajectories. We present a novel training framework using epilogue terminal rewards for learning specific behaviors from pre-computed trajectory data and demonstrate a successful transfer to hardware on the bipedal robot Cassie.

Optimizing Bipedal Maneuvers of Single Rigid-Body Models for Reinforcement Learning

In this work, we propose a method to generate reduced-order model reference trajectories for general classes of highly dynamic maneuvers for bipedal robots for use in sim-to-real reinforcement learning. Our approach is to utilize a single rigid-body model (SRBM) to optimize libraries of trajectories offline to be used as expert references in the reward function of a learned policy. This method translates the model's dynamically rich rotational and translational behaviour to a full-order robot model and successfully transfers to real hardware. The SRBM's simplicity allows for fast iteration and refinement of behaviors, while the robustness of learning-based controllers allows for highly dynamic motions to be transferred to hardware. % Within this work we introduce a set of transferability constraints that amend the SRBM dynamics to actual bipedal robot hardware, our framework for creating optimal trajectories for dynamic stepping, turning maneuvers and jumps as well as our approach to integrating reference trajectories to a reinforcement learning policy. Within this work we introduce a set of transferability constraints that amend the SRBM dynamics to actual bipedal robot hardware, our framework for creating optimal trajectories for a variety of highly dynamic maneuvers as well as our approach to integrating reference trajectories for a high-speed running reinforcement learning policy. We validate our methods on the bipedal robot Cassie on which we were successfully able to demonstrate highly dynamic grounded running gaits up to 3.0 m/s.

Beyond Value: CHECKLIST for Testing Inferences in Planning-Based RL

Reinforcement learning (RL) agents are commonly evaluated via their expected value over a distribution of test scenarios. Unfortunately, this evaluation approach provides limited evidence for post-deployment generalization beyond the test distribution. In this paper, we address this limitation by extending the recent CheckList testing methodology from natural language processing to planning-based RL. Specifically, we consider testing RL agents that make decisions via online tree search using a learned transition model and value function. The key idea is to improve the assessment of future performance via a CheckList approach for exploring and assessing the agent's inferences during tree search. The approach provides the user with an interface and general query-rule mechanism for identifying potential inference flaws and validating expected inference invariances. We present a user study involving knowledgeable AI researchers using the approach to evaluate an agent trained to play a complex real-time strategy game. The results show the approach is effective in allowing users to identify previously-unknown flaws in the agent's reasoning. In addition, our analysis provides insight into how AI experts use this type of testing approach, which may help improve future instantiations.