Organizations: National Key Laboratory of General Artificial Intelligence, BIGAI · Beijing Normal University · Communication University of China · Shanghai Jiao Tong University · B
Training locomotion policies for complex unstructured terrain requires a curriculum to avoid early exploration failures. However, since unstructured terrain lacks explicit difficulty ordering for curriculum design, existing methods resort to heuristic curricula over parameterized terrains. This abstraction limits generalization, as policies can overadapt to near-fixed perceptual patterns. To address this, we propose \ourname{}, an Trajectory-level Automatic Curriculum Learning framework that generates training tasks directly from unstructured terrain maps. At each curriculum update, the evaluator learns a difficulty function for the current policy that maps a given trajectory task to a difficulty score. The sampler then proposes new trajectories guided by the learned evaluator as the curriculum for the next policy update. This forms a closed loop in which the curriculum is iteratively matched to the evolving policy. Quantitative and qualitative experiments show that \ourname{} continuously provides effective curricula on unstructured terrain, improving trajectory success rate by 56.3% over direct training without curriculum. Compared with handcrafted curriculum learning, our method improves success rate by 18.5% on the hardest terrain tasks and by up to 39.74% when evaluating traversal from diverse approach directions on the same obstacle type.
Figures & tables
Figure 2: Framework overview. (a) The policy learns to follow waypoint trajectories, where a trajectory T is converted into velocity commands ct for policy training. (b) The current policy is rolled out on randomly sampled trajectories to collect transition descriptors ξi and success/failure labels yi . The CVAE encoder qϕ maps each descriptor to a latent code zi , and the evaluator predicts its policy-conditioned difficulty di . (c) Guided by the updated evaluator, the sampler uses MH-MCMC to generate capability-matched trajectories by sampling for transitions whose predicted difficulty di matches the target difficulty dtar . The generated trajectories are used for the next policy update, closing the curriculum loop. Fire and snowflake icons indicate training and inference modes.
Figure 3: Obstacle families in Map_HCL . From left to right: steps, gaps, parkour, and hurdles, each shown at two difficulty levels. Red lines and points denote the predefined waypoint trajectories used in the handcrafted curriculum.
Figure 4: Sampler-generated curriculum by our TACL on Map_A . (a) The terrain height field. (b)–(d) Trajectories sampled by our method at different policy-training epochs.
Figure 5: (a) Our method TACL adapts the training distribution to the evolving policy capability on Map_A . (b) Our method TACL achieves comparable results to the handcrafted curriculum baseline on major obstacle families in Map_HCL , without using predefined training trajectories.
Figure 6: Platform-climbing evaluation across approach directions. Each value is the average success rate over diverse approach directions, where our method shows more robust traversal than HCL.
Method
Eval. Acc.
Avg. Energy
Success Rate
RandST
–
–
35.2%
TACL w/o CVAE
67.3%
0.013
47.4%
TACL w/ CVAE
92.5%
0.021
94.6%
Table 1: Ablation of CVAE-based transition encoding on Map_A .