A Robotic Testbed for Autonomous Dump Pocket Cleaning Using Imitation Learning

Abstract

Dump pocket blockages at primary crushers cause production downtime and require manual clearing in hazardous restricted areas, with documented engulfment fatalities in hoppers and crusher zones. Autonomous approaches for dump pocket cleaning and blocked-crusher clearing remain limited, and the feasibility of state-of-the-art imitation learning (IL) for this excavation task remains largely unexplored. This paper introduces an experimental testbed and benchmark for evaluating IL architectures on autonomous dump pocket cleaning under controlled laboratory conditions. We benchmark four IL architectures: Action Chunking with Transformers (ACT), Diffusion Policy, and Vision-Language-Action models (π₀.₅ and SmolVLA), using a low-cost SO-ARM100 platform ($250) with granular bentonite material. In a preliminary single-session evaluation (10 minutes per model), π₀.₅ achieves the highest removal rate at 404 g, reaching 65% of the expert teleoperation estimate, followed by ACT (169 g), SmolVLA (113 g), and Diffusion Policy (57 g). π₀.₅'s lead is consistent with potential advantages from larger model capacity (3.7B vs. 450M parameters) and broader pretraining across diverse robot embodiments and tasks. ACT, trained from scratch on the same dataset, outperforms SmolVLA in this single-session benchmark, raising the hypothesis that for narrow single-task settings where the target domain differs from pretraining data, learning from scratch may remain competitive with fine-tuning a pretrained model. Diffusion Policy removes the least material, consistent with its visibly slower reactive behavior during rollout. This work establishes an open dataset (162 demonstrations) and a reproducible testbed to support future research on autonomous tasks for the mining industry.

Key Results

Single 10-minute autonomous session per model. Expert baseline estimated from demonstration data (62.04 g/min × 10 min). N=1 session per model — values represent single observed measurements.

Model	Removed (g / 10 min)	Rate (g/min)	% of Expert (est.)
Expert Teleoperation (est.)	≈ 620	62.0	100%
π₀.₅ (VLA)	404	40.4	65%
ACT	169	16.9	27%
SmolVLA	113	11.3	18%
Diffusion Policy	57	5.7	9%

These are exploratory observations pending systematic replication. Cross-architecture comparison is based on real-robot removed mass, not validation loss, because the models optimize different training objectives.

Scope, Limitations, and Transfer

The benchmark demonstrates laboratory feasibility in a simplified dump-pocket analogue, not mine-ready deployment. The testbed uses bentonite rather than heterogeneous ore, operates at roughly 1000× smaller scale than industrial hoppers (0.016 vs. 10–50 m³), and does not include dust, poor lighting, vibration, or mine-scale sensing constraints.

Scaling to industrial embodiments also requires new demonstrations on the target hardware. The algorithmic workflow may transfer, and the current dataset may support benchmarking or pretraining, but final deployment data must be collected at the relevant scale and hardware. This lab-to-mine data and embodiment gap is the primary limitation.

Testbed Contribution

The project provides a lab-scale dump-pocket testbed built on the open-source SO-ARM100 low-cost robot platform, rather than a proprietary arm design. The contribution is the mining task formulation, dataset, benchmark protocol, custom bucket setup, and reproducible evaluation pipeline for comparing imitation learning methods in a dump-pocket cleaning analogue.

Success Rollouts

One representative autonomous run per model (10-minute session, muted). Filmed with the overhead and wrist cameras used during evaluation.

π₀.₅ VLA · ~3.7B params

404 g removed · 40.4 g/min · 65% of expert

ACT No pretraining

169 g removed · 16.9 g/min · 27% of expert

SmolVLA VLA · 450M params

113 g removed · 11.3 g/min · 18% of expert

Diffusion Policy No pretraining

57 g removed · 5.7 g/min · 9% of expert

Figures

System overview — **Figure 1.** Experimental setup for autonomous dump pocket cleaning using imitation learning. Top left: teleoperated dataset (162 demonstrations). Top right: physical testbed with dual RGB cameras (top-view, wrist-mounted), dump pocket, and SO-ARM100 platform (follower and leader arms). Bottom: imitation learning pipeline receiving observations (camera streams, task prompt*, proprioceptive state), processing through policy network, and executing actions in closed loop. *Task prompt used only by VLA models (π₀.₅, SmolVLA) for language conditioning; ACT and Diffusion operate on vision and proprioception only.

**Figure 2.** Benchmark results and offline training metrics. (a) Validation loss trajectories; ACT is plotted as (val/loss)² for visual scale comparability — ACT uses a composite L₁+λKL objective while other models use MSE/flow-matching losses. Best checkpoints (stars): ACT 0.0754 @ 9,205 steps; π₀.₅ 0.0745 @ 25,774 steps; SmolVLA 0.0254 @ 14,728 steps; Diffusion 0.0141 @ 7,252 steps. This panel is a convergence diagnostic; cross-architecture comparison uses real-robot results only. (b) Total material removed (g) in a single 10-minute autonomous session. Expert baseline (~620 g) estimated from demonstration data (62.0 g/min, 2542 g over 40.99 min). π₀.₅ removes 404 g, followed by ACT (169 g), SmolVLA (113 g), and Diffusion Policy (57 g).

Resources

Paper & Code

Dataset

Model Checkpoints

Citation

@inproceedings{meza2026autonomous,
  title={A Robotic Testbed for Autonomous Dump Pocket Cleaning Using Imitation Learning},
  author={Meza Pinedo, Brik Henrry and Pajares Correa, Brian},
  booktitle={World Mining Congress 2026},
  year={2026}
}

Acknowledgments

We thank NONHUMAN for providing laboratory facilities and infrastructure support that enabled this research. We are grateful to the open-source robotics community, particularly the developers of the SO-ARM100 platform and the open-source robot learning library, whose accessible tools and collaborative spirit made these experiments possible. We also acknowledge PUCP for institutional support.