An autonomous walking excavator showing “unprecedented” levels of autonomy has been successfully demonstrated by ETH Zurich university.
The Hydraulic Excavator for an Autonomous Purpose (HEAP) was developed by Robotic Systems Lab, part of the public research university in Zurich.
In an open access paper published in the September issue of Automation in Construction, the team behind HEAP’s development say the excavator has successfully completed embankment excavations and the assembly of dry-stone walls.
HEAP specs
HEAP, which is based on a commercially available Menzi Muck M545 walking excavator, comes equipped with various sensors, including a Leica iCON iXE3 and two LiDAR sensors for perception.
To achieve active chassis balancing, all 14 stock hydraulic cylinders in the chassis were exchanged with new cylinders with an integrated control module that allows for precise control of cylinder position and force.
The paper says the chassis controller guarantees a stable stance of the machine for high-precision excavation.
HEAP also features a standard excavator bucket and a two-fingered gripper.
The test
In the tests, HEAP was able to create a two-faced embankment with 0.03m average error, as well as a more challenging s-curved embankment (30 m3 of soil displaced) with 0.05m average error.
As well as a standard excavator bucket, HEAP also features a two-fingered gripper, which was used to build dry-stone walls from stones with irregular shapes.
The machine’s two-fingered gripper was used to build dry-stone walls from stones with irregular shapes.
The stones are mapped by the machine itself, which picks them up and moves them in front of its roof-mounted LiDARs.
A task-specific planner provides grasping poses for the stones and a final pose in the wall with a feasible approach trajectory.
HEAP built a 4m tall, 10m long, and 2m wide wall with 109 stones from a collection of 131 scanned stones.
The stones used comprised natural stones and concrete blocks and had an average weight of 1,030kg.
The machine places the stones with a mean error of 0.128 m.
Summing up the HEAP test, the paper concludes: “In these unique autonomous missions, HEAP has shown unseen end-effector accuracy and unprecedented manipulation skills, e.g., in excavating free-form 3D landscapes or accurately placing large rocks to build a wall.”
It adds: “Although autonomous excavators already show higher absolute accuracy in many tasks compared to human operators, there are still things it cannot do as well.
“When a task needs constant adaptation and shows many unexpected events, a human operator can still reason and leverage his experience to solve the problem where a robot would simply be stuck.
“Repetitive and well-controlled tasks, e.g. digging a many kilometre-long pipeline trench through a desert, will be autonomously executed by robots soon, but as the research in this article is developed further, autonomous walking robots will begin to be able to tackle more and more sophisticated excavation and construction projects.”
Main image: Two HEAP excavators
Read next: Estonian contech start-up raises £1.2m in seed funding from UK-based investment firm
Are you a building professional? Sign up for a FREE MEMBERSHIP to upload news stories, post job vacancies, and connect with colleagues on our secure social feed.