Open issues in fixed-point-transformation-based adaptive control of an automatic convoy of vehicles: convergence assumptions, scalability, communication robustness, safety distances, and experimental validation

A 2007 INES conference paper proposed an adaptive control scheme for an automatic convoy (platoon) of four vehicles: the leading vehicle's commanded acceleration is deformed each control cycle by one step of a parametric fixed-point iteration, so that the last, traffic-facing vehicle tracks a smooth nominal trajectory through an unknown, saturating, monotone cascade of vehicle responses, avoiding dissipative inter-vehicle damping. Local convergence is proven only under a differentiable, strictly monotone, stationary response function, together with sign and magnitude conditions on the design parameters and the existence of the fixed point, and the scheme was validated only in two idealized four-vehicle simulations with instantaneous, loss-free sharing of acceleration data. The paper's own future-work agenda asks whether the scheme's claimed capabilities survive outside this envelope: under violated or relaxed convergence assumptions, with local tracking policies simpler than the PD-type kinematic feedback, for longer convoys, under communication delays and packet losses, with rigorously reduced safe inter-vehicle distances, and on real vehicles.

Background

An automatic convoy (also called a platoon or virtual train) is a string of vehicles that follow one another at short, automatically maintained distances. Convoy control has two classical aims: the motion of the last vehicle - the only part of the formation directly visible to following traffic - should be smooth, and energy should be saved by avoiding dissipative (viscous, braking-type) inter-vehicle forces. In a leader-follower setup each vehicle applies a local feedback controller regulating its distance to its predecessor; because actuators and tire forces saturate, the map from a vehicle's commanded acceleration to its realized acceleration is well modeled by a bounded, S-shaped (sigmoid-like) function, and the composition of these saturations along the chain makes the overall input-output behavior an unknown nonlinearity.

The 2007 INES paper "Adaptive Control of an Automatic Convoy of Vehicles" (DOI 10.1109/ines.2007.4283666) studied a four-vehicle chain of this kind. Each vehicle's local controller is a PD-type kinematic feedback, i.e., a proportional-derivative law acting like a spring-damper coupling on the tracking error; the leading vehicle's commanded acceleration $x$ is the control input; the last vehicle's realized acceleration is the output; and the composition of the saturated responses defines an unknown, strictly increasing response function $f$ (so $f'(x)>0$).

A fixed point of a map $g$ is a value left unchanged by it, $g(x_\star)=x_\star$; if $g$ is a contraction near $x_\star$ (it shrinks distances, $|g'(x_\star)|\le K<1$), Banach's fixed-point theorem guarantees that the iterates $x_{n+1}=g(x_n)$ converge to $x_\star$. The paper's adaptive idea is to avoid both energy-dissipating strong damping and model-based adaptive control (which would require an analytical model of the saturated cascade) by deforming the commanded input, once per control cycle, with the parametric fixed-point iteration

$g\left(x \mid x^{d}, D_{-}, \Delta_{+}\right) := \frac{\left(f(x)-\Delta_{+}\right)\left(x-D_{-}\right)}{x^{d}-\Delta_{+}} + D_{-}$

where $x^{d}$ is the desired last-vehicle acceleration (computed by the PD-type feedback on the last vehicle's tracking error) and $D_{-}$, $\Delta_{+}$ are design parameters (the simulations used $D_{-}=-10\ \mathrm{m/s^{2}}$ and $\Delta_{+}=20\ \mathrm{m/s^{2}}$). Any $x_\star$ with $f(x_\star)=x^{d}$ is a fixed point of $g$; differentiating gives $g'(x_\star)=1+f'(x_\star)\frac{x_\star-D_{-}}{x^{d}-\Delta_{+}}$, and the paper proves local convergence by a contraction (Cauchy-sequence) argument under sign and magnitude conditions on $D_{-}$, $\Delta_{+}$, $x_\star$, $x^{d}$, and $f'(x_\star)$, assuming that $f$ is differentiable and monotone increasing and that the fixed point exists. In two four-vehicle simulations (nominal gaps $L_i=5$ m; saturation amplitudes $A_2=30$, $A_3=20$, $A_4=10\ \mathrm{m/s^{2}}$; and a harder setting with stiffnesses $P_2=0.09$, $P_3=0.08$, $P_4=0.06\ \mathrm{s^{-2}}$ and saturations $A_2=9$, $A_3=8$, $A_4=7\ \mathrm{m/s^{2}}$), the last vehicle's tracking error dropped from order $10^{-1}$ m (non-adaptive, visibly vibrating) to order $10^{-4}$ m without re-tuning the adaptive parameters.

Everything validated sits inside one envelope: four vehicles, one specific PD-type tracking policy, perfect and instantaneous sharing of acceleration data between vehicles, an exactly monotone, differentiable, stationary response function, and simulation only. The question below is the paper's own future-work agenda, asking whether the promised capabilities - smooth traffic-facing motion without dissipative damping, robustness without re-tuning, energy savings, and possibly reduced safe tracking/braking distances between convoy members - survive outside that envelope. For orientation: the same research group later developed the fixed-point-deformation idea into a method family called robust fixed point transformations (RFPT), and the modern platooning literature treats convoys with linear, model-predictive, and learning-based designs.

Problem Statement

The problem: for the fixed-point-transformation-based adaptive convoy control scheme of the 2007 INES paper - leading-vehicle commanded acceleration deformed each control cycle by one step of the parametric fixed-point iteration $g\left(x \mid x^{d}, D_{-}, \Delta_{+}\right)$ defined in the background, desired response generated by a PD-type tracking policy on the last vehicle's error, and the last vehicle's realized acceleration communicated back to the controller - establish which of the scheme's claimed capabilities hold outside the validated four-vehicle, delay-free, simulation-only envelope. The source poses six linked facets of this single objective of characterizing the scheme's validity domain:

1. Convergence assumptions: does the local convergence of the iteration survive violations or relaxations of the convergence hypotheses of the source proof - the differentiability, strict monotonicity, and stationarity of the response function $f$, together with the sign and magnitude conditions on the design parameters and the existence of the fixed point - for example under more complex vehicle dynamics, external disturbances, non-monotone or only piecewise-smooth responses, or a desired response $x^{d}$ that changes within the adaptation?
2. Tracking policy: does the adaptive compensation still improve tracking when the local controllers use a simpler tracking policy than the PD-type kinematic feedback, as the source suggests studying? The answer must define and justify the class of simpler local policies it tests and compare the outcome against the PD-type kinematic feedback.
3. Scale: how does the scheme behave for convoys longer than the simulated four vehicles, including the motion of the intermediate vehicles?
4. Communication: is the scheme robust to delays and packet losses in the inter-vehicle sharing of acceleration data on which the adaptation relies?
5. Safety: can the safe tracking/braking distances between convoy members be rigorously reduced under the adaptive scheme, with a quantitative safety-margin analysis?
6. Validation: does the scheme work on real vehicles rather than only in simulation? This facet is resolved by evidence from real vehicles or by a refutation; a simulation-only study does not by itself resolve it.

For each regime considered, the question is which of the scheme's claimed benefits - smooth last-vehicle motion without dissipative damping, robustness without re-tuning, shorter safe headways - hold and which are lost. The facets are kept unified because they share the same plant, the same adaptive mechanism, and the same objective; the source poses them as one future-work agenda for this scheme.

References

1. J.K. Tar, "Adaptive Control of an Automatic Convoy of Vehicles", in: 2007 11th International Conference on Intelligent Engineering Systems (INES), IEEE, 2007, pp. 21-26. DOI: 10.1109/ines.2007.4283666, URL: https://doi.org/10.1109/ines.2007.4283666.
2. J.K. Tar, "Adaptive Control of a Semi-Automatic Convoy of Unmodeled Internal Degrees of Freedom", in: 2007 5th International Symposium on Intelligent Systems and Informatics (SISY), IEEE, 2007, pp. 129-134. DOI: 10.1109/sisy.2007.4342638, URL: https://doi.org/10.1109/sisy.2007.4342638.
3. J.K. Tar, I.J. Rudas, J.F. Bito, J.A. Tenreiro Machado, K. Kozlowski, "Adaptive VS/SM controller based on robust fixed point transformations", in: 2009 International Conference on Intelligent Engineering Systems (INES), IEEE, 2009, pp. 51-55. DOI: 10.1109/ines.2009.4924737, URL: https://doi.org/10.1109/ines.2009.4924737.
4. J.K. Tar, I.J. Rudas, J.F. Bito, J.A. Tenreiro Machado, K. Kozlowski, "Adaptive controller for systems of fractional dynamics based on robust fixed point transformations", in: 2009 7th International Symposium on Applied Machine Intelligence and Informatics (SAMI), IEEE, 2009, pp. 117-123. DOI: 10.1109/sami.2009.4956622, URL: https://doi.org/10.1109/sami.2009.4956622.
5. Hemza Redjimi, Jozsef K. Tar, "On the effects of time-delay on precision degradation in fixed point transformation-based adaptive control", in: 2017 IEEE 30th Neumann Colloquium (NC), IEEE, 2017, pp. 000125-000130. DOI: 10.1109/NC.2017.8263267, URL: https://doi.org/10.1109/NC.2017.8263267.
6. Shulin Feng, Wei Wang, Huanshui Zhang, "Three-dimensional pursuer convoy by using guidance laws", Journal of Control Theory and Applications, 2013, pp. 442-453. DOI: 10.1007/s11768-013-2012-3, URL: https://doi.org/10.1007/s11768-013-2012-3.
7. Kazunori Urayama, Kenichiro Nonaka, Kazuma Sekiguchi, "An inter-vehicular distance control considering path tracking and attitude angular error for obstacle avoidance", in: 2016 IEEE/SICE International Symposium on System Integration (SII), IEEE, 2016, pp. 774-779. DOI: 10.1109/SII.2016.7844093, URL: https://doi.org/10.1109/SII.2016.7844093.
8. Tamas Becsi, Adam Szabo, Balint Kovari, Szilard Aradi, Peter Gaspar, "Reinforcement Learning Based Control Design for a Floating Piston Pneumatic Gearbox Actuator", IEEE Access, vol. 8, 2020, pp. 147295-147312. DOI: 10.1109/ACCESS.2020.3015576, URL: https://doi.org/10.1109/ACCESS.2020.3015576.
9. Hakan Köroğlu, "String-Stable Cooperative Adaptive Cruise Control With Minimized Time Headway in the Face of Delayed Communication", IEEE Control Systems Letters, vol. 8, 2024, pp. 400-405. DOI: 10.1109/LCSYS.2024.3392716, URL: https://doi.org/10.1109/LCSYS.2024.3392716.
10. Namya Bagree, Charles Noren, Damanpreet Singh, Matthew Travers, Bhaskar Vundurthy, "Distributed Optimal Control Framework for High-Speed Convoys: Theory and Hardware Results", arXiv:2211.06287, 2022. DOI: 10.48550/arXiv.2211.06287, URL: https://arxiv.org/abs/2211.06287.
11. Owen Burns, Hossein Maghsoumi, Yaser Fallah, Israel Charles, "OpenConvoy: Universal Platform for Real-World Testing of Cooperative Driving Systems", arXiv:2405.18600, 2024. DOI: 10.48550/arXiv.2405.18600, URL: https://arxiv.org/abs/2405.18600.