Driving · 2026-11-10
DreamLander: A 6-DOF World Model based Reinforcement Learning Planetary Lander
George Nehma, Patrick Quinn, Madhur Tiwari, Camilo Riano-Rios
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
DreamLander is a DreamerV3-based model-based reinforcement learning method for a realistic 6-DOF entry descent landing problem. Built as an end-to-end capable guidance/control framework, it uses a latent world model (RSSM) to learn dynamics and train an actor-critic policy, with state sensor observations (Ray-Cast height mapping, IMU velocity/angular velocity, and leg contact). The authors report landing within 2 meters of the designated landing zone and velocity norm < 0.3 m/s.
Why it matters
The paper targets safety-critical, fuel-limited planetary landing guidance/control. By combining a DreamerV3 world model with latent imagination and an actor-critic controller, DreamLander aims to provide sample efficiency and generalization across EDL scenarios, while incorporating realistic sensor observations rather than using only ground-truth states. The authors position it as progress toward end-to-end Entry Descent Landing and discuss how the world model could support safety/stability guarantees.
Method
- DreamLander uses DreamerV3 world-model RL: an autoencoder maps observations to a latent space, an RSSM learns/predicts dynamics in latent space, and an actor-critic policy outputs actions from latent representations.
- Training uses latent imagination: the model creates imagined latent future states and uses them to further train the actor-critic policy.
- The simulated environment is a 6-DOF EDL setup built in IsaacLab/PhysX, with state sensor observations including Ray-Cast height mapping, IMU velocity and angular velocity (with stated noise levels), and leg contact sensing. The policy can be used end-to-end or its guidance tracked by controllers (e.g., PID/adaptive/MPC).
Abstract (from arXiv)
Soft powered descent landings are becoming evermore important in the field of space exploration and scientific advancement, because of their ability to deliver mission critical payloads, both highly accurately and safely. The high efficiency restriction on planetary landings due to the severely limited fuel supply results in the need for optimal guidance trajectories and control. Model based reinforcement learning is proving to be an important tool in simplifying the highly complex process of generating optimal guidance trajectories and control. In this work, we present DreamLander, a state of the art model based reinforcement learning algorithm based on DreamerV3, to generate optimal guidance and control for a realistic 6 degree of freedom entry descent landing problem. The model based approach creates a sample efficient and highly generalizable model that can be applied to a variety of scenarios whilst maintaining a high degree of accuracy, landing within 2 meters of the designated landing zone, and with velocity norm less than 0.3 meters per second. DreamLander makes use of state sensor observations, making progress toward End-to-End solutions of the Entry Descent Landing problem.
Related papers
- Mastering Diverse Domains through World Models
- DiffDreamer: Coupling World Models with Differentiable Simulation for Agile Quadrotor Control
- Mastering Atari with Discrete World Models
- DayDreamer: World Models for Physical Robot Learning
- Explore Where It Matters: Targeted Exploration for Efficient Rover Adaptation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-11-09 · Indexed by Pub-AI: 2026-11-10 UTC
DreamLander: A 6-DOF World Model based Reinforcement Learning Planetary Lander
Preprint posted: 2026-11-09 · Original record ↗
- arxiv
- 2611.12824
Explicit source versions (1)
- v1 · 2026-11-09
- Preprint available
Source date: 2026-11-09 · First detected: 2026-11-10 UTC · Historical baseline
Source evidence (2)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.12824v1; arxiv:announce_type=new; item pubDate (official metadata)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Soft powered descent landings are becoming evermore important in the field of space exploration and scientific advancement, because of their ability to deliver mission critical payloads, both highly accurately and safely. The high efficiency restriction on planetary landings due to the severely limited fuel supply results in the need for optimal guidance trajectories and control. Model based reinforcement learning is proving to be an important tool in simplifying the highly complex process of generating optimal guidance trajectories and control. In this work, we present DreamLander, a state of the art model based reinforcement learning algorithm based on DreamerV3, to generate optimal guidance and control for a realistic 6 degree of freedom entry descent landing problem. The model based approach creates a sample efficient and highly generalizable model that can be applied to a variety of scenarios whilst maintaining a high degree of accuracy, landing within 2 meters of the designated landing zone, and with velocity norm less than 0.3 meters per second. DreamLander makes use of state sensor observations, making progress toward End-to-End solutions of the Entry Descent Landing problem.
