Source-linked AI summary
On Shaky Ground: Characterising Non-Rigid Base Effects in Mobile Manipulation
Zongyuan Zhang, Christopher Lehnert, Will N. Browne, Jonathan M. Roberts
TL;DR
Lightweight mobile bases can move or vibrate under contact, but how this affects force and tool motion during sustained contact is insufficiently characterised. The paper combines exploratory simulation with sanding experiments on two lightweight wheeled testbeds, jointly measuring force, tool motion, and vibration. Sustained contact still produced substantial force shortfall and tool error at low resistance; increasing resistance improved force realisation and arrested further displacement.
Problem
How base motion manifests jointly in realised contact force and tool motion during sustained contact on lightweight mobile manipulators remains insufficiently characterised.
Source: §1 · p. 1, §2.3 · p. 3Method
The study combines exploratory simulation with sanding experiments on two lightweight platforms, measuring contact force, tool motion, and base vibration together.
Source: §1 · p. 2, §2.3 · p. 3Results
At low resistance, only 54–70% of a 100 N command was realised with TCP errors of about 100 mm; higher resistance raised realised force to 88–101% and arrested further TCP-error growth.
Source: §7 · p. 9Takeaways & Limitations
Force, tool-motion, and vibration measurements provide complementary views of non-rigid base behaviour during contact-task execution.
Source: §6 · p. 9Each condition was run once, and measured tool motion combines manipulator tracking error, arm deflection, and base motion.
Source: §6 · p. 9
Abstract
from arXiv · showhide
Mobile robot manipulators extend contact-task capability beyond the workspace of fixed robot installations, but their bases must also sustain the forces generated by the task. On lightweight platforms, base motion and support compliance can change the force and motion realised at the workpiece. How these non-rigid base effects manifest during sustained contact remains insufficiently characterised. This paper investigates non-rigid base effects in mobile robot manipulators through a sanding case study combining exploratory simulation with experiments on two lightweight, unactuated wheeled testbeds. Joint measurements of contact force, tool motion, and base vibration show that sustained contact coexisted with substantial task error: under low base-ground resistance, only 54-70% of a 100 N command was realised, with tool-centre-point errors reaching 100 mm. Increasing base-ground resistance during contact improved force realisation and arrested further tool displacement, while base vibration did not vary consistently with force-tracking error in the tested conditions. These observations illustrate how non-rigid base behaviour can affect contact-task execution and highlight the value of examining force, motion, and vibration together when characterising lightweight mobile manipulation.
1 Introduction
Mobile manipulators expand contact-task access beyond fixed workspaces, but lightweight bases may move or vibrate under contact loads. This study characterises those effects through sanding experiments and simulation, finding force shortfall and tool error during sustained contact, with resistance changes improving force realisation and arresting further displacement.
1 Introduction
- Fixed-base manipulators offer accurate motion and force control but restrict work to a fixed workspace, whereas mobile bases extend access between work areas.Mobile systems are attractive for large or distributed structures such as aircraft, ships, and building surfaces.Source: §1 · p. 1
- Contact loads pass through the manipulator and base to the ground, where limited traction and interface compliance can produce base motion and vibration.Base motion can contribute to TCP error and absorb commanded tool advance, leaving force below the command even while contact is maintained.Source: §1 · pp. 1–2
- Under low base–ground resistance, sustained contact coexisted with force shortfall and TCP error, while increasing resistance improved force realisation and arrested further tool displacement.The assessment used two lightweight platforms and varied commanded force, arm posture, tool excitation, and support compliance.Source: §1 · p. 2
- Joint force, tool-motion, and base-vibration measurements identify responses that future active compensation strategies would need to address.The measurements provide a complementary view of contact execution rather than relying on force tracking alone.Source: §1 · p. 2
2 Related Work
Prior work addresses manipulator-side finishing dynamics, mobile machining stability, and base compliance, but sustained force-controlled contact on lightweight stationary platforms remains under-characterised. This study measures contact force, tool motion, and base vibration together in that setting.
2.1 Robotic Surface Finishing
- Finishing research addresses vibration and compliance in the manipulator, tool, or workpiece, while mobile-base interaction with the ground has received less attention.Prior approaches include feed-speed and force control, posture selection, compliant end-effectors, and tool-holder design.Source: §2.1 · p. 2
2.2 Mobile Machining and Base Stability
- Mobile machining systems use heavy bases or structured environments to recover stability, which reduces the mobility and deployment ease sought in lightweight platforms.Stability analyses and wheel-slip studies provide relevant foundations, but do not resolve the sustained-contact characterisation gap.Source: §2.2 · pp. 2–3
2.3 Base Compliance and Physical Interaction
- Base-compliance studies establish effects on physical interaction, but joint measurements of realised force and tool motion remain limited for sustained contact on lightweight stationary platforms.The present study measures contact force, tool motion, and base vibration together.Source: §2.3 · p. 3
3 Exploratory Simulation
An exploratory simulation of a mobile manipulator on a rigid-wheel base was used to identify possible contact failure modes and inform the physical experiments. Sweeps revealed oscillation, slip, and loss of stability, but the simulated trends are qualitative rather than quantitative hardware predictions.
3 Exploratory Simulation
- The exploratory simulation was used to identify potential failure modes, examine parameter effects, and inform the physical experiments.Its role was exploratory rather than predictive.Source: §3 · p. 3
3.1 Environment and Control
- The MuJoCo setup models a UR10 on a differential-drive base with rigid wheels, traction and stability limits, a compliant sanding pad, and hybrid force–position control.It varies contact force, wheel friction, arm posture, mass distribution, and tool excitation; wheel–ground compliance is not modeled.Source: §3.1 · p. 3
3.2 Observed Failure Modes
- Parameter sweeps produced tool-and-base oscillation, wheel slip, and loss of stability under contact loading.Oscillation increased with friction, commanded normal force, and arm extension; slip became more likely with larger loads, lower wheel–ground friction, and postures unloading driven wheels; extended postures and unfavourable mass distributions reduced stability margin.Source: §3.2 · p. 3
- The simulated trends are qualitative because contact parameters were not identified from the platforms and the rigid-wheel model omits tyre, platform, and support-surface compliance.The simulation therefore identifies plausible failure modes and qualitative trends rather than predicting hardware results quantitatively.Source: §3.2 · p. 3
4 Physical Experiments
The physical study uses two lightweight wheeled platforms to examine how resistance to base motion and support compliance affect force-controlled sanding. It combines force, tool-motion, and vibration measurements across three experiments and evaluates them during steady contact.
4 Physical Experiments
- Two lightweight platforms with contrasting support conditions are used to examine whether non-rigid base effects persist across platforms.The cart and trolley are tested on the same force-controlled sanding task; both are unactuated surrogates for lightweight mobile-robot bases.Source: §4 · p. 3
4.1 Platforms
- The cart uses an aluminium frame and deck, while the lighter trolley uses a plywood deck with a compliant foam layer beneath it.The foam provides a repeatable compliant support representing tyres, suspension, or soft ground.Source: §4.1 · p. 4
4.2 Experimental Setup
- Base–ground resistance is varied by leaving the wheels free or locking them, while support compliance and tool speed are also controlled.Locking the wheels increases resistance to base motion but does not rigidly fix the platform; force and vibration sensors, plus camera-tracked tool markers, measure the response.Source: §4.2 · p. 4
4.3 Force Controller
- A task-space admittance controller converts normal-force error into end-effector velocity while maintaining tangential position.Cartesian commands are resolved to joint velocities using a quadratic programme.Source: §4.3 · p. 4
4.4 Task and Trials
- Three experiments vary base resistance, posture, commanded force, tool speed, and support compliance, including within-trial resistance switches during contact.Trials compare low and high resistance on the cart and trolley; each trial is performed once, so results are treated as a case series. TCP error is measured relative to the workpiece, although the tangential target is held relative to the base.Source: §4.4 · pp. 4–5
4.5 Evaluation Metrics
- Force, tool-motion, and vibration metrics are evaluated over steady contact, excluding contact-onset, switching, and retraction transients.For resistance-switch trials, the two phases are evaluated separately; base vibration is peak-to-peak deck-normal acceleration over 1 s intervals, summarized by its median and maximum.Source: §4.5 · p. 5
5 Results
Across the trials, sustained contact could coexist with force shortfall and tool motion at low base resistance, while increasing resistance improved force realisation and arrested further tool displacement. Vibration varied with tool, support, and contact conditions, and the simulation provided a qualitative comparison with hardware behaviour.
5.1 Force Shortfall and TCP Motion at Low Resistance
- At 100 N, low-resistance trials realised 54–70% of command, while TCP error reached 98–103 mm in T5/T6 and 96 mm on the second platform.The shortfall persisted through steady contact; in T5 and T6, TCP error accumulated until resistance was increased part-way through contact.Source: §5.1 · p. 6, Figure 4 · p. 6, Figure 6 · p. 8
5.3 Dependence on Posture and Commanded Force
- At 100 N, lowering posture raised the realised ratio from 0.59 to 0.68, while reducing command to 25 N yielded 26.3 N realised force and 1.9 N error.The posture change reduced but did not remove the sustained shortfall; at 25 N, force tracking was close.Source: §5.3 · p. 6, Figure 4 · p. 6
5.2 Response to Increased Base Resistance
- After resistance increased, TCP-error growth stopped within about 1 s and force error fell 66% in T5 and 75% in T6.Force ratios reached 0.88 in T5 and 1.01 in T6 over the high-resistance intervals.Source: §5.2 · p. 6, Figure 5 · p. 7
5.4 Vibration Across Tool and Support Conditions
- Median vibration rose in every low-to-high resistance comparison, yet nearly ninefold tool-speed differences did not increase force variation; support and contact also changed vibration.At low resistance and a 20 N command, vibration was 16 m s−2 on the hard floor and 42 m s−2 on foam.Source: §5.4 · pp. 7–8, Figure 7 · p. 8
5.5 Relation to the Simulation
- Simulation and hardware both showed sensitivity to base–ground interaction and posture, but hardware support compliance was absent from the rigid-wheel model; adverse simulated stability-loss conditions were not reproduced for safety.The table gives a qualitative comparison; loss of stability was observed in simulation, not in the physical trials.Source: Table 3 · p. 7, §5.5 · p. 8
6 Discussion
The discussion argues that sustained contact alone did not ensure commanded force or workpiece-relative tool position. It treats force, tool motion, and vibration as complementary measurements, while bounding the findings by the case-series scope.
Discussion
- At low resistance under a 100 N command, contact persisted while force remained below command and the tool moved about 100 mm; increasing resistance arrested TCP-error growth and raised realised force.The same direction of change was observed on the second platform.Source: §6 · p. 8
- Base motion absorbing part of the commanded tool advance is consistent with the observations; posture changes the contact-force moment about wheel contacts, and load changes the support reaction.The posture and load comparisons show how task conditions affect the expression of non-rigid base effects.Source: §6 · pp. 8–9
- Force, the tool marker, and base vibration serve complementary purposes: they describe command tracking, workpiece-relative tool motion, and responses to excitation and support conditions.The discussion notes that high resistance improved force tracking despite increased vibration.Source: §6 · p. 9
- Each condition was run once, and the study covers two platforms and a limited range of lightweight configurations; surface finish and material removal were not assessed.The measured tool motion also includes manipulator tracking error, arm deflection, and base motion.Source: §6 · p. 9
7 Conclusion and Future Work
On two lightweight wheeled testbeds, low base–ground resistance coincided with substantial force shortfall and TCP error during sanding. Increasing resistance during contact improved force realisation and arrested further tool displacement, while responses also varied with task conditions and platform dynamics.
7 Conclusion and Future Work
- At low base–ground resistance, a 100 N command yielded 54–70% realised force and TCP errors of about 100 mm in camera-tracked trials.During high-resistance phases, average realised force rose to 88–101% of command, and further TCP-error growth stopped within about 1 s.Source: §7 · p. 9
- A 25 N command was closely tracked at low resistance, while a low posture reduced the force shortfall.Separate low- and high-resistance trials on the second platform showed the same direction of change as the within-trial comparisons.Source: §7 · p. 9
- Base vibration depended distinctly on tool excitation, support, and contact loading; hardware exposed compliance-related responses absent from the rigid-wheel simulation model.The simulation provided a qualitative framework for base motion, oscillation, and stability.Source: §7 · p. 9
- Future work should vary force, posture, and support independently, test more platforms, and relate dynamic response to surface finish and material removal.Actuated mobile-robot studies should assess coordinated wheel-motor and arm commands against available traction and base stability.Source: §7 · p. 9