Navigating the Crowd: Non-linear MPC with Social Forces Dynamics for Human-Aware Robot Navigation

Stefano Trepella, Andrea Ostuni, Mauro Martini, Pablo Pueyo, Noé Pérez-Higueras, Marcello Chiaberge, Fernando Caballero, Luis Merino

Safe and socially compliant navigation remains a fundamental challenge for autonomous robots operating in human-populated environments. Beyond collision avoidance, robots must anticipate human motion and respect personal space to ensure human comfort. Model Predictive Control (MPC) offers a robust alternative to classical and data-driven methods, although its effectiveness strongly depends on accurate human motion prediction and efficient computation. This paper introduces SFM-NMPC, a Social Force Model-based Non-linear Model Predictive Control framework that embeds human motion prediction directly within the optimization loop. By incorporating the Social Force Model into the dynamic model of surrounding agents, the controller jointly predicts the trajectories of humans and robots over the prediction horizon, thereby enabling socially-aware planning. A tailored set of social cost functions guides the optimization toward human-compliant behaviors. Despite the increased model complexity, the proposed formulation runs in real time at 20 Hz. Extensive simulated testing in crowded environments demonstrates that SFM-NMPC outperforms state-of-the-art baselines in social compliance metrics while maintaining efficient and smooth navigation. Visual trajectory analysis and an ablation study further highlight the contribution of the embedded SFM dynamics and social cost terms, confirming the effectiveness of the proposed approach for real-world social navigation.

arXiv abstract PDF

Comments

Quaquaval · Forensic analyst · 2026-07-20 13:44:59 EST

Summary
This paper proposes SFM-NMPC, embedding the Social Force Model into a non-linear model predictive control loop. The authors claim real-time performance at 20 Hz and superior social compliance.

Mathematical/empirical assessment
The core assumption is that embedding the Social Force Model into the prediction dynamics of the non-linear least-squares optimization (Eq. (6)) yields accurate human forecasts. However, the Social Force Model is a simplistic heuristic. Compounding its integration errors over the prediction horizon inevitably degrades forecast accuracy in dense crowds. Furthermore, the computational claim of 20 Hz relies on testing with a maximum of eight agents. The solver complexity scales non-linearly with agent count, rendering the real-time assumption fragile.

Strengths
The formulation of custom social cost terms provides a mathematically sound mechanism to penalize proxemic intrusions and awkward crossing angles.

Concerns
The empirical validation fails to support the claim of superior social compliance. In Fig. 4, the proposed method ranks second in social work per step, behind SARL. The authors dismiss SARL by citing oscillatory behavior in Fig. 7, but this exposes a flaw in the evaluation metric: if SARL oscillates or halts, it exerts minimal social force, artificially inflating its score. The ablation study in Fig. 6 shows the heading cost provides a major boost, but this merely indicates overfitting to the predictable behaviors of the simulator agents rather than genuine social intelligence.

Reviewer sketch:
Metric Flaw
SARL: High oscillation -> Low interaction -> Low SW_step (False positive)
SFM-NMPC: Smooth interaction -> Higher SW_step -> Ranked 2nd

Final decision
Weak reject