1. Why Send a Robot Back and Forth on a Forest Road?

In Japan's forests, deer stripping bark and grazing on undergrowth are serious and worsening problems. It is well known that deer flee when they sense a human presence on a forest road — but sending a person out every single day is simply not realistic.

"If we can get a robot to travel back and forth along a forest road while broadcasting human voices, deer should learn to avoid that area," Kamoda says.

Patrolling alone is not enough, however. If a person has to retrieve and recharge the robot every time the battery dies, there is no point in automating the patrol. The robot must return to its station on its own, charge itself, and head back out — an "outdoor Roomba" is exactly the ultimate goal of this research.

2. It Started with a Gasoline Brush Cutter — Over Two Years of Trial and Error

Kamoda began this research in the final year of his undergraduate degree, before entering graduate school. At the time, he bought an off-the-shelf gasoline-powered brush cutter and converted it into a robot.

After all that trial and error, the conclusion was clear: "Electric drive is absolutely the way to go." Kamoda raised his hand to take over a CuboRex crawler unit (CuGoMEGA) that had been sitting unused in the lab.

Even after switching to the electric crawler, things did not go smoothly. A system that worked perfectly on the university campus fell apart completely when brought out to a real forest road.

3. Satellite-Free "Map Matching" Achieves Under-3 cm Accuracy

The biggest obstacle on a forest road is self-localization. Dense forest canopy blocks GNSS (GPS) signals, making it impossible to rely on satellites for position data. The system therefore uses only a LiDAR (Light Detection and Ranging) sensor to scan the surrounding environment and navigate.

Simply driving with LiDAR was not enough, however. Because the goal is to repeat the same route, the conventional "LiDAR SLAM" approach — which builds a map and estimates position simultaneously while moving — tends to lose track of the current position on vibration-prone terrain such as forest roads. To address this, LiDAR point clouds are matched against a pre-built map, and IMU (accelerometer and gyroscope) data is fused with the 3D LiDAR data to realize a vibration-resistant map-matching method.

Once each morning, the robot is driven manually by remote control to scan the surrounding environment and build a 3D map. During the afternoon's autonomous runs, the robot navigates by matching that pre-built map against real-time sensor data.

4. 100% Docking Success — The Secret Behind a Mechanism That Snaps Into Place on Rough Terrain

Auto-docking is just as challenging as autonomous navigation. On a forest road where ground conditions change every run, the robot must return precisely to its station and connect to the charging connector — all automatically.

The final segment of the route is set as a straight line aimed directly at the station. The robot simply follows that path to approach. At the last moment, a magnet mounted on the station side is detected by the robot's magnetic sensor, which triggers a "docking complete" judgment and stops the robot.

5. "The Terrain Capability Is Incredible" — An Honest Field Review of the CuGoMEGA M1

At the end of the interview, we asked Kamoda what it was actually like to use the CuGoMEGA M1 in the field. Having struggled with a gasoline-powered machine, his words carry real weight.

When we mentioned that the next-generation model, the CuGoMEGA M2, is now available, Kamoda's eyes lit up.

6. What's Next — Farmland Pest Control, Continued Through a 3–4 Year Doctoral Program

Kamoda has plans to push the research even further. The next step is to expand the scope from forest roads to agricultural fields. Farmland offers few natural landmarks, making LiDAR-only localization difficult; GNSS will need to be added to the autonomous navigation stack.

To accelerate the research, Kamoda has enrolled in the doctoral program.

The quest to build a robot that keeps moving autonomously outdoors has only just begun.