NavBench
收藏资源简介:
NavBench是一个多领域基准数据集,旨在为基于强化学习的自主导航策略的训练和评估提供统一的平台。它构建于IsaacLab之上,通过标准化任务定义,使不同机器人能够应对各种导航挑战,无需重新设计任务或自定义评估指标。数据集包括陆地、水上和太空等多个领域的机器人,并提供了多种导航任务,如到达指定位置、保持指定姿态、穿越一系列航点等。通过模拟到现实世界的验证,NavBench简化了可适应的基于强化学习的导航策略的开发。
NavBench is a multi-domain benchmark dataset designed to provide a unified platform for the training and evaluation of reinforcement learning-based autonomous navigation strategies. Built upon IsaacLab, it enables diverse robots to tackle various navigation challenges through standardized task definitions, eliminating the need for task redesign or custom evaluation metrics. The dataset includes robots across multiple domains such as land, water, and space, and offers a variety of navigation tasks including reaching a designated position, maintaining a specified attitude, traversing a sequence of waypoints, and more. Leveraging sim-to-real validation, NavBench simplifies the development of adaptable reinforcement learning-based navigation strategies.

- 1NavBench: A Unified Robotics Benchmark for Reinforcement Learning-Based Autonomous NavigationSnT, University of Luxembourg · 2025年



