Source-linked AI summary
Walking on Roofs: Exploring the Potential of Walking Robots for Construction Work on Roofs
Bjoern-Felix Dettmar, Arne Roennau
TL;DR
Roof work is hazardous and repetitive, but roof environments remain a largely unexplored robotics application. The paper evaluates a commercial quadruped on a roof test rig, finding that the unmodified Go2 cannot operate safely and reliably on roofs under the tested conditions.
Problem
Roof work combines high risks for human workers with repetitive, physically demanding tasks, yet roof environments remain a largely unexplored robotics application.
Source: §6 · p. 11Method
Results
Without appropriate adaptations, the Go2 was not capable of safe and reliable roof operation, with critical failures during incline walking at 25° and a quadratic relationship between slippage and incline.
Source: §6 · p. 11Takeaways & Limitations
Roof deployment will require co-design of terrain-aware control, perception-driven foothold selection, and environment-specific foot hardware.
Source: §5 · p. 10Tests used one non-high-end Go2 with its standard gait at one speed under ideal conditions.
Source: §5 · p. 11
Abstract
from arXiv · showhide
This paper investigates the feasibility of deploying quadruped walking robots for the automation of work in roof environments. While quadrupeds have demonstrated versatility across various domains, their large-scale deployment remains limited, partly due to lack of application-specific designs. Roof environments represent a novel and unexplored use case, combining high safety risks for human workers with repetitive, strenuous tasks that could benefit from robotic assistance. A dedicated test rig of a roof's surface was designed to evaluate the baseline performance of a commercial quadruped, the \emph{Unitree Go2}, in this new environment. Experiments revealed that standard ball feet are inherently inadequate for locomotion on sloped roofs: slippage increased quadratically with incline angle, get-up and lie-down sequences were only possible on small inclines, and critical failures already occurred regularly on moderate inclines of 25{\deg}. This work provides the first systematic assessment of quadruped locomotion in roof environments, highlighting both the potential and the current limitations of this application and establishes a baseline for future research. Effective solutions will require a combination of task-oriented foot designs, advanced contact mechanics and environment-specific control strategies, such as reinforcement learning for roof-adapted gaits.
1 Introduction
Roof work combines hazardous conditions with repetitive, physically demanding tasks, while roofs remain a largely unexplored robotics application. This section motivates roof-specific quadruped research and introduces the paper’s baseline-evaluation contributions.
1 Introduction
- Roof tasks expose workers to serious fall risks while requiring repetitive, physically demanding work.The roofing industry in Germany employed over 99,000 people across more than 13,500 companies in 2023.Source: §1 · p. 2
- Quadruped deployment remains limited despite advances in mobility and autonomy, motivating investigation of new applications with well-defined tasks and challenges.Proposed commercial applications have included inspection and maintenance, search and rescue, and military and surveillance.Source: §1 · p. 2
- Roof tiles combine moderate-to-steep slopes and surface irregularities that challenge locomotion, while their repeating structure offers an opportunity for environment-specific foot designs.Residential roof inclines commonly range from 25° to 60°.Source: §1 · p. 3
- The paper introduces a systematic roof-locomotion evaluation, quantitative failure thresholds for core skills, and motion-capture measurement of foot slippage.The evaluated skills include balancing, get-up/lie-down motions, and incline walking.Source: §1 · p. 3
2 Evaluating Quadruped Locomotion in Roof Environments
The section frames friction and stability as the key determinants of roof locomotion and motivates measuring slippage objectively. It also explains why standard round feet may be poorly suited to inclined, complex roof surfaces.
2 Evaluating Quadruped Locomotion in Roof Environments
- Roof locomotion is assessed primarily through friction, with loss of stability and falling marking the upper performance limit.The paper estimates friction from slippage on a defined incline and uses OptiTrack motion capture, calibrated to below 1.0 mm accuracy.Source: §2 · p. 3
- Standard round feet offer simple, robust contact but their limited area can increase stress and slippage on inclines, and their geometry does not adapt to varied terrain.The paper identifies customized, environment-adapted feet as a topic for future work.Source: §2 · p. 3
3 Experimental Setup
The experiments establish a repeatable Go2 baseline on an adjustable roof-tile rig. They measure friction, balance, get-up/lie-down performance, and incline-walking slippage under specified conditions.
3 Experimental Setup
- Dynamic friction was measured by dragging a Go2 ball foot over flat and curved roof-tile surfaces under a 60 N normal force, with ten repetitions per surface.The foot was dragged 100 mm; dynamic friction was used because deformation made static-friction measurement unreliable.Source: §3 · p. 4
- The adjustable rig used common German Frankfurther Pfanne tiles to test Go2 balance, get-up/lie-down sequences, and incline walking, with each experiment repeated ten times.Balance was tested facing uphill; get-up/lie-down success required completing both actions in uphill, downhill, and sideways orientations.Source: §3 · p. 4
- Incline walking used the standard gait at a target speed of 1.0 m s−1, with reflective foot markers tracking backward movement during stance as slippage.Only increments with vertical displacement below 1.0 mm were accumulated to exclude swing-phase movement.Source: §3 · p. 4
4 Baseline Results
The unmodified Go2 showed important limits on the roof rig: it failed below manufacturer-claimed climbing angles, and its walking slippage rose quadratically with incline. Static, get-up/lie-down, and walking tests exposed distinct failure conditions.
4 Baseline Results
- Without significant modifications, the Go2 was not suited for operation within roof environments, and its manufacturer-claimed slope-climbing capability of up to 40° was not achieved under the test conditions.Source: §4 · p. 5
- The ball-foot friction coefficient was 1.144±0.1299 on flat tile and 1.430±0.1298 on curved tile, corresponding to material-specific incline angles of (48.84 ± 3.22)° and (55.03 ± 2.44)°, respectively.Source: §4 · p. 5
- 27.58±0.72° was the average critical standing angle, with failures primarily following automatic limb repositioning apparently triggered by limb slippage or atypical joint angles.Source: §4 · p. 5
- Get-up and lie-down succeeded only up to 11.0° in head-up and head-down orientations and 7.0° sideways; failures occurred when the center of mass left the support polygon.Source: §4 · p. 5
- Across incline tests from 0° to 25°, frequent falls occurred at 25°; steeper slopes showed slippage, especially at the rear foot, as abrupt slip-offs rather than continuous sliding.Source: §4 · p. 5
- Incline increased trajectory variability and reduced traveled distance, while the roof’s tile-row steps could cause slipping even on flat ground.The curved tile areas also promoted lateral deviation; the reported regression found no explainable correlation between incline and trajectory deviation.Source: §4 · p. 7
- Slippage had a highly significant quadratic relationship with incline angle (p(F) < 0.001), while traveled distance had a highly significant linear relationship (p < 0.0035) and trajectory deviation had no explainable correlation.Fig. 4 presents the mean distance, slippage, and trajectory-deviation metrics with regression models.Source: §4 · p. 7, Figure 4 · p. 8
5 Discussion
Tests show that the unmodified Go2 is not suited to roof operation, with frequent critical failures during incline walking at 25°. The discussion links this performance to dynamic contact and control limits and calls for coordinated hardware and software adaptations.
5 Discussion
- Frequent critical failures during incline walking at 25° show that the Go2 is currently not suited for roof deployment.The authors evaluate balancing, get-up/lie-down sequences, and incline walking as the first baseline for quadrupeds in roof environments.Source: §5 · p. 8
- Front-foot slippage, rearward center-of-mass shift, unchanged push-off behavior, and undamped impacts help explain why measured friction overstates dynamic performance.The controlled friction values are theoretical maxima, and roof locomotion is limited by dynamic stability and control as well as material friction.Source: §5 · pp. 8–9
- Friction measurements alone cannot predict dynamic locomotion: transient load redistribution, impacts, and controller behavior also limit performance.Balancing failures were preceded by front-limb slippage as the center of mass shifted rearward and reduced front-foot normal forces.Source: §5 · p. 9
- Reliable roof locomotion calls for incline-aware control, terrain perception, adaptable high-friction feet, and terrain-aware get-up and lie-down motions.The discussion also identifies level activation areas as an alternative to incline-aware recovery behaviors.Source: §5 · p. 10
- A co-design approach should combine terrain-aware control, perception-driven foothold selection, and environment-specific foot hardware.Suggested investigations include roof-specific reinforcement learning and feet with larger, adaptable contact areas, high-friction materials, and actuated ankle joints.Source: §5 · p. 10
- The study used one commercial robot, its standard gait at 1.0 m s−1, and ideal conditions without debris, vegetation, weather, or payloads.The authors recommend testing different platforms, conditions, and settings for more general insights.Source: §5 · p. 11
6 Conclusion
Roof work combines hazards for people with repetitive, physically demanding tasks, motivating investigation of quadruped assistance. Tests on a roof rig found that the unmodified Go2 was not safe or reliable on inclines, motivating combined control, perception, and hardware adaptations.
6 Conclusion
- Roof work combines high-risk conditions for humans with repetitive, physically demanding tasks that could benefit from robotic assistance.Source: §6 · p. 11
- The unmodified Go2 reached critical failure before common residential roof inclines, while slippage increased quadratically with inclination and get-up/lie-down sequences were unsafe on inclines.The authors conclude that the robot is not capable of safe and reliable roof operation without appropriate adaptations.Source: §6 · p. 11
- Future work should combine tailored locomotion strategies and perception with hardware adaptations, including roof-specific gaits and actuated end-effectors.The authors also call for accurate modeling of friction and ground interactions.Source: §6 · p. 11