RoboSense: Leveraging Robotaxi Fleets as Drive-by Sensors for Urban Traffic Monitoring

Yilin Wang; Yiheng Feng

cs.ROSep 6, 2026

RoboSense: Leveraging Robotaxi Fleets as Drive-by Sensors for Urban Traffic Monitoring

Authors: Yilin Wang, Yiheng Feng

Organizations: Lyles School of Civil and Construction Engineering, Purdue University, West Lafayette, USA

Abstract

Urban traffic monitoring plays a critical role in safety analysis, congestion management, and incident response. The growing deployment of robotaxis creates a new opportunity for network-level traffic monitoring. Although robotaxis are primarily designed to serve passengers, they can also be leveraged as drive-by sensors to collect traffic data. Compared to conventional infrastructure sensors or probe vehicles, a fleet of robotaxis forms a cooperative perception environment, which can collectively gather spatially and temporally continuous traffic information. This paper proposes a novel dynamic robotaxi routing framework that explicitly incorporates traffic monitoring tasks as an objective. The framework introduces: (1) a cell-based network representation that aligns with sensing capabilities of robotaxis; (2) a cell-level monitoring metric to quantify spatiotemporal robotaxi coverage; and (3) a mixed-integer linear programming (MILP) formulation that jointly minimizes time-dependent travel time and maximizes traffic monitoring performance. A 5 by 5 urban grid network is built in SUMO to evaluate the framework under three robotaxi market penetration rates (2%, 5%, and 10%) with a range of objective weight combinations. Results show that incorporating spatiotemporal network coverage in the objective function can effectively improve the traffic monitoring performance. Interestingly, with appropriate weights between the two objectives, monitoring performance and robotaxi average speed can be improved simultaneously. This suggests better network monitoring leads to more accurate traffic state prediction and improved mobility. This win-win situation could incentivize robotaxi operators to contribute their vehicles as drive-by sensors for traffic monitoring.

Figures & tables

Panel of Figure 1 — Figure 1: An illustrative example on routing selection with different considerations

Figure 2: Data modalities in traffic monitoring

Panel of Figure 3 — Figure 3: Network CTM representation

Figure 4: Overview of the routing for traffic monitoring framework

Notation	Meaning
$A$	Set of robotaxis
$I$	Set of cells in the network
$E$	Set of edges in the network
$T$	Planning horizon
$x_{i,t}^{a}$	Binary variable, equals to 1 if robotaxi $a$ enters cell $i$ at time $t$ , and 0 otherwise
$z_{i,j,t,s}^{a}$	Binary variable, equals to 1 if robotaxi $a$ enters cell $i$ at time $t$ and leaves cell $i$ at time $s$ , heading to cell $j$ , and 0 otherwise

Table 1: Variables and notations

Figure 6: Relationship between $\omega$ and $x$

Panel of Figure 7 — Figure 7: Route choice and coverage under different weight parameters

Panel of Figure 9 — Figure 9: $5\times 5$ urban network for SUMO and CTM representation

Figure 11: Fixed-time signal timing plan for intersections

Origin	Destination	Flow Rate	Origin	Destination	Flow Rate
$A1$	$E5$	$700\ \mathrm{veh/h}$	$E3$	$A3$	$350\ \mathrm{veh/h}$
$A2$	$E4$	$350\ \mathrm{veh/h}$	$E2$	$A4$	$350\ \mathrm{veh/h}$
$A3$	$E3$	$350\ \mathrm{veh/h}$	$E1$	$A5$	$700\ \mathrm{veh/h}$
$A4$	$E2$	$350\ \mathrm{veh/h}$	$D1$	$B5$	$350\ \mathrm{veh/h}$
$A5$	$E1$	$700\ \mathrm{veh/h}$	$C1$	$C5$	$350\ \mathrm{veh/h}$
$B5$	$D1$	$350\ \mathrm{veh/h}$	$B1$	$D5$	$350\ \mathrm{veh/h}$

Table 2: Origin-destination (OD) demand for each node in the grid network

Figure 12: Calibrated fundamental diagram for cell transmission model

Panel of Figure 13 — Figure 13: Comparison between CTM result and ground truth for link $E120$

Figure 14: Solving time distribution under different MPRs

Case( $\alpha_{1}$ : $\alpha_{2}$ )	Penetration Rate	Robotaxi Avg. Travel Time (s)	Robotaxi Avg. Travel Distance (m)	Robotaxi Avg. Speed (m/s)	BV Avg. Travel Time (s)	CTM MAE	Network Coverage	Vehicle Coverage
$1:10$	$2\%$	767.58	3026.48	4.16	825.54	1.2508	3.29%	10.25%
$1:100$	$2\%$	776.08	3064.15	4.13	827.55	1.2319	3.32%	10.16%
$1:1000$	$2\%$	778.39	3074.90	4.14	820.80	1.2118	3.37%	10.55%
$1:1500$	$2\%$	790.23	3089.95	4.20	822.72	1.1548	3.43%	10.63%
$1:2000$	$2\%$	775.08	3085.89	4.19	822.95	1.1961	3.36%	10.35%
$1:2500$	$2\%$	770.88	3041.93	4.16	822.11	1.1573	3.40%	10.57%

Table 3: Experiment results under different robotaxi MPRs and objective weights

Figure 15: Robotaxi average travel time (left axis, solid lines with filled markers) and network coverage (right axis, dashed lines with hollow markers) under different routing strategies $\alpha_{1}:\alpha_{2}$ and MPRs. Blue, red, and green correspond to MPRs of $2\%$ , $5\%$ , and $10\%$ , respectively.

Panel of Figure 16 — Figure 16: Sensing power curves over the coverage weighting parameter $\alpha_{2}$ under four different settings. Top-left: MPR $=2\%$ ; top-right: MPR $=5\%$ ; bottom: MPR $=10\%$ .

Panel of Figure 17 — Figure 17: Segment popularity ( $p_{i}$ ) under different coverage weights across MPRs. (a)–(f): 2% MPR; (g)–(l): 5% MPR; (m)–(r): 10% MPR. Columns from left to right correspond to $\alpha_{2}=10,100,1000,3000,10000,10^{5}$ .

Panel of Figure 18 — Figure 18: Robotaxi travel distance distributions under different coverage weights across MPRs. Each subplot compares the distribution at a given $\alpha_{2}$ (red) against the most travel-time-oriented case $\alpha_{2}=10$ (blue). (a)–(f): 2% MPR; (g)–(l): 5% MPR; (m)–(r): 10% MPR; (s)–(x): 10% MPR without de-routing budget. Columns from left to right correspond to $\alpha_{2}=100,1000,3000,5000,10000,10^{5}$ .

Panel of Figure 19 — Figure 19: Robotaxi travel time distributions under different coverage weights across MPRs. The layout matches Figure 18 . (a)–(f): 2% MPR; (g)–(l): 5% MPR; (m)–(r): 10% MPR; (s)–(t): 10% MPR without de-routing budget. The rightward shift of travel time distributions reflects both extended travel distances and reduced travel speeds at high $\alpha_{2}$ values.

Figure 20: (a) Diversification ratio $U(\alpha_{2})/U(\alpha_{2}{=}10)$ ; (b) Relative average route length $\bar{L}(\alpha_{2})/\bar{L}(\alpha_{2}{=}10)$ . Both panels report values relative to the $\alpha_{2}{=}10$ baseline across MPR settings and the no-budget variant.

Panel of Figure 21 — Figure 21: Robotaxi average travel time (a) and average travel distance (b), each together with the fleet coverage, under the vehicle weighted objective.

Case $\alpha_{1}$ : $\alpha_{2}$	MPR	Avg. TT (s)	Avg. Dist. (m)	Network Coverage	Vehicle Coverage	Case $\alpha_{1}$ : $\alpha_{2}$	MPR	Avg. TT (s)	Avg. Dist. (m)	Network Coverage	Vehicle Coverage
$1:10$	$2\%$	768.49	3022.73	3.23%	11.53%	$1:10$	$10\%$	770.17	3063.14	14.18%	46.18%
$1:30$	$2\%$	779.61	2997.98	3.33%	12.11%	$1:30$	$10\%$	778.91	3096.80	14.54%	47.15%
$1:60$	$2\%$	777.68	3012.53	3.24%	11.81%	$1:60$	$10\%$	762.86	3095.49	15.01%	48.47%
$1:100$	$2\%$	771.60	3033.14	3.37%	12.38%	$1:100$	$10\%$	776.34	3126.29	15.58%	50.05%
$1:200$	$2\%$	787.15	3014.03	3.38%	12.62%	$1:200$	$10\%$	833.15	3389.07	16.69%	53.65%
$1:300$	$2\%$	824.70	2989.19	3.52%	13.57%	$1:300$	$10\%$	879.43	3563.31	17.26%	54.60%

Table 4: Experiment results under different robotaxi MPRs and objective weights with revised objective function

Panel of Figure 22 — Figure 22: Average observed vehicle count per time step under (a) cell based coverage objective, and (b) vehicle weighted coverage objective.

Intelligent Transportation Systems

Intelligent Transportation Systems apply sensing, communication, AI, and optimization to make transport networks safer and more efficient.

See all 303 papers

Prediction is Better than Detection: Traffic Congestion Control using Drones
Sep 30, 2026Samira Hayat, Christian RaffelsbergerMulti-Robot Systems Traffic Signal Control
Lakeside Labs GmbH Klagenfurt, Austria
Monitoring Urban Traffic Dynamics at Fine Spatiotemporal Resolution Using Distributed Acoustic Sensing and Deep Learning
Sep 23, 2026Hao Tian, Heng Cai, Xiaowei Chen +1Intelligent Transportation Systems Traffic Flow Estimation
Department of Geography, Texas A&M University, College Station, TX 77840, USA · Department of Geology and Geophysics, Texas A&M University, College Station, TX 77840, USA
Synergizing Drone Delivery Order Pooling and Road Network Monitoring through Monitoring-Task Orderization
Oct 4, 2026Yulong Hu, Meng Xu, Sen Li +1Aerial Robotics Intelligent Transportation Systems
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong, China · School of Architecture, Civil and Environmental Engineering (ENAC), École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland

Sep 30, 2026J/K move · Enter open · S save

Prediction is Better than Detection: Traffic Congestion Control using Drones

Samira Hayat, Christian Raffelsberger

Lakeside Labs GmbH Klagenfurt, Austria

Multi-Robot Systems Traffic Signal Control Traffic Forecasting

Open paper AskPDF arXiv

A central question in deploying teams of mobile robots for persistent monitoring is how task performance scales with fleet size, and whether this scaling holds once sensing drives downstream action rather than mere observation. We study this question for a team of drones performing traffic-jam detection and prediction in a simulated road network, whose reports drive an adaptive traffic-signal controller in closed loop. We build a multi-agent simulation, with vehicles following Nagel-Schreckenberg cellular-automaton dynamics and drones patrolling junctions via a round-robin policy, and sweep fleet size, traffic level, and network size to evaluate detection rate, detection delay, and prediction rate. We show how performance plateaus for fleet size approximating the number of junctions being monitored, and offer a general fleet-provisioning rule for persistent-monitoring deployments. More significantly, adapting the signal on a predicted jam, rather than a detected one, roughly doubles the resulting reduction in jam duration, showing that the value of onboard prediction in a sensing-to-action pipeline can exceed the value of adding more robots. Prediction accuracy, not sensing coverage, is now the binding constraint on further improvement, pointing to onboard inference, not fleet size, as the more promising direction for future work.

Abstract

Intelligent Transportation Systems

Explore similar work

Prediction is Better than Detection: Traffic Congestion Control using Drones

Monitoring Urban Traffic Dynamics at Fine Spatiotemporal Resolution Using Distributed Acoustic Sensing and Deep Learning

Synergizing Drone Delivery Order Pooling and Road Network Monitoring through Monitoring-Task Orderization