Digiforest workshop
At a workshop in the Forest of Dean under grey February skies, Digiforest Oxford team members demonstrated aspects of the robotic navigation and semantic mapping they had been working on. They showcased the quadruped mobile robot (an adapted ANYmal) and a LiDAR sensor fixed to a backpack.
Digiforest LiDAR backpack demonstration in the Forest of Dean
Unlike most robots deployed in industrial settings, Digiforest’s ANYmal prototype has been trained to traverse rougher forest terrain. The waterproof robot sits lower to the ground and walks slowly in a move to improve its stability. It is fitted with a LiDAR sensor and various cameras. Designed to work wirelessly, the robot can operate for around 1.5 hours, depending on the terrain. On average, it covers one metre of ground per second.
Map of the forest created by ANYmal's LiDAR sensing
On their computers, the researchers set routes for the robot to follow. The robot walks these routes, adapting them slightly when obstacles (namely trees ) are encountered. As the robot moves, the LiDAR sensor collects data on the diameter of the tree trunks passed, enabling biomass to be calculated. The LiDAR creates a map of the forest floor and offers a below-canopy view of the forest to complement aerial scans. It is hoped that eventually the robot will be programmed to provide species ID for the trees, using bark to register identity. Information on species is vital to help foresters calculate carbon .
The robot’s movement is undeniably uncanny. When stationary, it sways slightly, its body rising and falling in a heaving movement reminiscent of breathing. This swaying stems from the robot’s programming to align itself with gravity. Shifts in the surrounding environment mean the robot is constantly adjusting its stance.
While the robot might be used to measure biomass, monitor environmental change over time (by comparing annual scans), and assess unsafe or inaccessible areas, its current design presents challenges. Moreover, the robot’s deployment raises important social, ecological and political questions.
Firstly, the robot has limited agility and finds it challenging to navigate obstacles such as tree stumps, suggesting it would struggle to proceed if it met with tangled understory or thickly planted trees. The robot’s interactions with the trees in its pathway are also somewhat laborious - a process of collision, then slow course correction. The Digiforest team concede that if the robot falls over it will struggle to right itself. The biomass measuring process can also be of limited accuracy. Among the trunks verified using a tape measure, some have a 2-4 inch margin of error. The robotic equipment is expensive, cumbersome, and heavy to transport, while the large number of datapoints in the forest scans results in slow uploads. Finally, the cameras struggle to incorporate colour as the lighting behind the trees often produces overexposed images.
Digiforest's adapted ANYmal prototype
The social, ecological and political considerations raised by the robot include further questions of how species, from deer to dogs to humans, will interact with the robot. The robot may also have implications for certain forestry jobs. While unadapted ANYmal robots like this currently cost around £100,000, with time these robots may be seen as a cost-effective alternative to human labour . Their use could also centralise decision-making, and remote actors may also reduce the possibilities for local communities to participate. Finally, the robot offers a limited view of the forest, failing to capture complex ecological systems, soil carbon or understory.
The LiDAR backpack was a simpler mechanism, performing the same sensing tasks as the robot's LiDAR, but was worn by a human. The wearer must walk slowly and steadily to enable the LiDAR to create a scan of the surrounding trees. While this backpack allowed more agile movement through the trees and did not so directly challenge human participation in the forest, the issues with accurate tree trunk measurements persisted.
Digiforest's adapted ANYmal prototype
The Oxford Digiforest team is working on improving volume estimation, incorporating species ID, and enhancing drone speed.
As robotics begins to be deployed more frequently within forestry, their social, ecological and political implications warrant further scrutiny.
