Key Takeaways
- ERNEST (Exploration Rover for Navigating Extreme Sloped Terrain) improves upon the classic rocker‑bogey suspension by adding active joints that let the rover shift weight, steer in any direction, and switch between energy‑saving passive mode and terrain‑capable active mode.
- The rover was prototyped in three generations, starting with small 2‑ft testbeds to evaluate 11 active‑suspension configurations in a lunar‑regolith simulant trailer before scaling to a full‑size platform with a mast‑mounted head.
- Autonomy was achieved through reinforcement‑learning training in a high‑fidelity virtual environment fed with real‑hardware data, allowing ERNEST to navigate obstacle courses in JPL’s Mars Yard without human joystick input.
- Ongoing work aims to fuse active‑suspension decision‑making with long‑range intelligent path planning, enabling the rover to select optimal gaits, avoid hazards, and tackle steep slopes on future Moon or Mars missions.
- Development began in 2022 with JPL internal R&D funding and is now supported by NASA’s Mars Exploration Program and the Exploration Science Strategy and Integration Office; Caltech manages JPL for NASA.
Overview and Mechanical Design
The ERNEST project set out to evolve the proven rocker‑bogey suspension that has served every Mars rover since Sojourner. While the passive rocker‑bogey keeps roughly equal load on all six wheels through pivots and struts, the ERNEST team hypothesized that adding actuation could yield better performance on extreme slopes and irregular terrain. Their solution incorporates two powered joints at the front that articulate a gimbal, enabling the rover to modify its posture and weight distribution actively. This active suspension can be disengaged via a clutch mechanism, reverting to the simpler, more energy‑efficient passive mode when the terrain permits. The result is a hybrid system that retains the reliability of rocker‑bogey while gaining the adaptability needed for steep, uneven surfaces.
Steering and Mobility Modes
Beyond suspension, ERNEST features four steerable wheels, granting it omnidirectional capability—including sideways crabbing, diagonal traversal, and precise pivot turns. The active suspension works in concert with this steering to produce specialized gaits such as “squirming” (a wave‑like motion that lifts and sets wheels sequentially), “wheel‑walking” (where wheels lift and step over obstacles while the body remains relatively level), and traditional obstacle‑climbing. By adjusting which wheels are powered, steered, or left passive, the rover can select the most efficient gait for a given obstacle, reducing slip and minimizing the risk of becoming immobilized on loose regolith or sharp rocks.
Prototyping Journey
Before arriving at the final design, the team built two early prototypes, each roughly two feet (0.6 m) long, to explore eleven distinct active‑suspension layouts. These testbeds were placed in a trailer filled with lunar‑regolith simulant, allowing the engineers to run systematic experiments across a range of slope angles and surface conditions over several months. Data from these trials informed decisions about actuator placement, gear ratios, and control logic, ultimately converging on a configuration that balanced mobility improvement with mechanical simplicity and mass constraints.
Scaling Up and Hardware Integration
Having validated the concept at small scale, the team scaled ERNEST to a full‑size rover. The upgraded vehicle features a rectangular head mounted on a 4.5‑foot‑tall (1.4 m) mast, which houses sensors, communication equipment, and a computational unit for autonomy. The mechanical structure was completed in September 2024, but at that stage the rover still required a human operator using a joystick to issue movement commands. This manual phase was essential for verifying that all actuators, sensors, and power systems functioned correctly before entrusting the rover with self‑direction.
Training Autonomy with Reinforcement Learning
To transition from tele‑operation to true autonomy, the ERNEST team turned to reinforcement learning (RL), a machine‑learning paradigm where an agent learns optimal behavior through trial‑and‑error interactions with its environment. The Dynamics and Real‑Time Simulation Laboratory at JPL constructed a high‑fidelity virtual replica of ERNEST, complete with accurate dynamics, sensor noise, and terrain physics. Engineers logged the rover’s real‑world responses to varied terrains—loose sand, rocky outcrops, steep inclines—and fed this data into the simulator. Running thousands of parallel simulation episodes on a high‑performance computing cluster, the team accumulated the equivalent of many months of real‑world testing in just a few weekends, allowing the RL policy to discover effective strategies for weight distribution, gait selection, and obstacle avoidance.
Field Validation in the Mars Yard
After the virtual training phase, the team tested ERNEST’s autonomous algorithms in JPL’s Mars Yard, an outdoor terrain facility designed to mimic Martian surface conditions. They constructed an obstacle course containing sand ripples, rubble piles, step‑like obstacles, and steep slopes that would typically stall a passive‑suspension rover. Upon releasing the rover with its learned policy, ERNEST successfully navigated the course unaided, demonstrating its ability to perceive terrain, choose appropriate gaits, and switch between active and passive suspension as needed. Repeated runs showed consistent performance, building confidence that the autonomy system could cope with the variability expected on actual planetary surfaces.
Future Autonomy Enhancements
Buoyed by these results, Hari Nayar’s group is launching a new project that integrates ERNEST’s real‑time suspension decision‑making with longer‑range intelligent navigation. The objective is to enable the rover to construct efficient, safe paths that not only circumvent hazardous zones (e.g., deep loose sand, large boulders) but also deliberately engage surmountable obstacles using the most suitable active‑suspension gait. By combining high‑level route planning with low‑level motion control, the rover could maximize travel distance while minimizing energy expenditure and risk—a critical capability for extended missions on the Moon’s rugged highlands or Mars’ dissected terrains.
Funding, Management, and Timeline
Work on ERNEST commenced in 2022, initially financed by JPL’s internal research and development allocation. As the technology matured, support expanded to include NASA’s Mars Exploration Program and the Exploration Science Strategy and Integration Office within the Science Mission Directorate at NASA Headquarters in Washington, D.C. The Jet Propulsion Laboratory, managed for NASA by the California Institute of Technology (Caltech) in Pasadena, California, continues to oversee the project’s execution, ensuring that technical milestones align with broader planetary exploration goals. The ongoing development reflects a coordinated effort to mature a versatile mobility platform that could serve as a workhorse for future lunar and Martian surface missions.