Cost Model¶
CostNav evaluates navigation policies through a comprehensive economic cost-revenue model grounded in real-world data. All parameters are referenced from industry sources including SEC filings, AIS injury reports, manufacturer specs, and delivery platform pricing.
For the full derivation, see Section 3.1 of the paper and the cost formula spreadsheet.
Cost Framework Overview¶
The total profit of a navigation policy is:
where \(\mathbb{R}\) is cumulative delivery revenue, CAPEX is upfront capital expenditure, OPEX is per-delivery operational cost, and \(N\) is the number of completed deliveries.
The contribution margin per delivery is:
where \(R = P_{\text{MktRobotDeli}} \times S_{\text{EpisodeTermSLA}}\) (base delivery fee x SLA compliance; zero if timeout).
The Break-Even Point (BEP) is the number of deliveries required to recover capital expenditure:
When contribution margin is negative (\(R < C_{\text{OPEX}}\)), the BEP is undefined --- the system is not economically viable.
Capital Expenditure (CAPEX)¶
One-time fixed investments incurred before deployment.
Hardware Cost¶
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Robot Cost (\(P_{\text{Robot}}\)) | 8,000 | $/robot | Retail price, Segway E1 delivery robot [21] |
| LiDAR Cost (\(P_{\text{Lidar}}\)) | 3,000 | $/robot | Livox LiDAR for delivery robots [28] |
| GPS Cost (\(P_{\text{GPS}}\)) | 2,000 | $/robot | RTK Device (with IMU) for delivery robots [28] |
Sensor Configuration
Rule-based methods (Nav2) use Robot + LiDAR + GPS = $13,000. Learning-based methods use Robot only = $8,000 (CANVAS adds GPS = $10,000).
Data Collection Cost¶
For learning-based navigation, the cost of collecting training data:
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Data Collector Wage (\(P_{\text{DataCollector}}\)) | 33 | $/hour | U.S. median hourly wage for data collection personnel [26] |
| Data Collection Time | 65.63 | hours | Total teleoperation hours in CostNav |
CAPEX Summary (Table 3 from paper)¶
| Component | Nav2 w/ AMCL | Nav2 w/ GPS | GNM | ViNT | NoMaD | NavDP | CANVAS |
|---|---|---|---|---|---|---|---|
| Hardware | 13,000 | 13,000 | 8,000 | 8,000 | 8,000 | 8,000 | 10,000 |
| Data Collection | 0 | 0 | 2,165.79 | 2,165.79 | 2,165.79 | 2,165.79 | 2,165.79 |
| Total CAPEX | 13,000 | 13,000 | 10,165.79 | 10,165.79 | 10,165.79 | 10,165.79 | 12,165.79 |
Operational Expenditure (OPEX)¶
Variable costs incurred on a per-delivery basis, directly influenced by navigation behavior. Organized into three categories: direct costs (electricity, repair), service costs (customer compensation), and liability costs (pedestrian and property damage).
Electricity Cost¶
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Electricity Rate (\(P_{\text{Elec}}\)) | 0.2704 | $/kWh | U.S. average retail electricity price [48] |
| Electro-Mechanical Efficiency (\(C_{\text{ElectroMechanicalEff}}\)) | 0.73 | unitless | Derived: battery roundtrip (0.95) x motor (0.82) x inverter (0.99) x charge (0.95) [4], [41], [42], [56] |
Repair Cost¶
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Useful Life of Robots (\(T_{\text{RobotLife}}\)) | 2 | years | Expected operational lifespan per SEC filings [30] |
| Useful Delivery Run Count (\(N_{\text{RobotLifeRun}}\)) | 18,250 | run/robot | Derived: \(T_{\text{RobotLife}} \times 365 \times N_{\text{Delivery}}\) |
| Repair Rate (\(C_{\text{Repair}}\)) | 10--20 | %/year | Annual maintenance cost as % of robot price [13], [15], [36] |
Service Compensation Cost¶
Failed deliveries incur compensation for customer refunds:
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Market Robot Delivery Price (\(P_{\text{MktRobotDeli}}\)) | 3.49 | $/run | Consumer delivery fee for campus robot services [47] |
| Avg Market Food Price (\(P_{\text{MktFood}}\)) | 31.93 | $/run | Derived: \(P_{\text{Refund}} - P_{\text{MktRobotDeli}}\) |
| Avg Refund Price (\(P_{\text{Refund}}\)) | 35.42 | $/run | Industry average order value for delivery failures [43] |
Pedestrian Safety Cost¶
Modeled using the Abbreviated Injury Scale (AIS), measured from collision delta-v:
where \(P(\text{AIS} \mid \Delta v)\) is the probability of AIS severity given impact speed [51], and \(K_{\text{InjuryAdjustment}}\) adjusts for vehicle weight differences between the delivery robot and the crash report reference vehicle.
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Avg Pedestrian Damage (AIS0) | 380 | $/event | Economic cost per MAIS pedestrian incident [3], [22] |
| Avg Pedestrian Damage (AIS1) | 8,487 | $/event | Same as above |
| Avg Pedestrian Damage (AIS2) | 60,464 | $/event | Same as above |
| Avg Pedestrian Damage (AIS3) | 261,200 | $/event | Same as above |
| MAIS Report Passenger Vehicle Weight | 4,536 | kg | Derived: 10,000 lbs = 4,536 kg, per MAIS Injury Report [51] |
Property Damage Cost¶
Collisions with urban infrastructure incur repair or replacement costs:
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Mailbox Damage Cost | 50 | $/event | Municipal replacement cost [45] |
| Trash Bin Damage Cost | 50 | $/event | Municipal replacement cost [7] |
| Building Glass Damage Cost | 300 | $/event | Estimated repair cost [54] |
| Bollard Damage Cost | 65 | $/event | Estimated replacement cost [52] |
OPEX Summary (Table 3 from paper, $/run)¶
| Component | Nav2 w/ AMCL | Nav2 w/ GPS | GNM | ViNT | NoMaD | NavDP | CANVAS |
|---|---|---|---|---|---|---|---|
| Electricity | 0.0029 | 0.0031 | 0.0005 | 0.0009 | 0.0003 | 0.0001 | 0.0027 |
| Service Compensation | 5.4021 | 10.1322 | 3.4900 | 3.1410 | 3.4551 | 3.4900 | 3.8910 |
| Pedestrian Safety | 29.3200 | 20.5800 | 11.6400 | 29.8900 | 9.7600 | 9.3000 | 14.3800 |
| Property Damage | 0.0000 | 1.0000 | 3.6500 | 6.0000 | 12.0000 | 0.0000 | 6.0000 |
| Repair | 10.8493 | 7.2329 | 5.2603 | 11.8356 | 6.9041 | 2.9589 | 6.5753 |
| Total OPEX | 45.5743 | 38.9482 | 24.0407 | 50.8675 | 32.1195 | 15.7491 | 30.8490 |
Revenue¶
Revenue per delivery is the market delivery fee ($3.49), modulated by SLA compliance. Deliveries exceeding the timeout receive zero revenue. Food intactness determines service completeness.
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Delivery Runs per Day (\(N_{\text{Delivery}}\)) | 25 | runs/day | Reported daily throughput per robot, SEC filings [29] |
| Delivery Failure Rate (\(C_{\text{DeliveryFailure}}\)) | 0.06 | %/run | Operational failure rate per SEC filings [29] |
| Operator:Robot Ratio (\(N_{\text{Fleet}}\)) | 4 | robot/operator | Level 4 autonomy supervision ratio per SEC filings [29] |
| Robot Operator Wage (\(P_{\text{Operator}}\)) | 24 | $/hour | U.S. median hourly wage for robot operators [27] |
Simulation-Measured Inputs¶
These values come from the Isaac Sim physics simulation and feed into the cost formulas above:
| Parameter | Value | Unit | Reference |
|---|---|---|---|
| Robot Speed (\(C_{\text{RobotSpeed}}\)) | 6.00 | km/h | Maximum operational velocity, Segway E1 spec [40] |
| Robot Max Speed (\(C_{\text{RobotMaxSpeed}}\)) | 12.80 | km/h | Maximum nominal speed of Segway E1 [28] |
| Robot Weight (\(C_{\text{RobotWeight}}\)) | 50 | kg | Maximum off-load weight, Segway E1 [28] |
| Avg Delivery Time (\(T_{\text{AvgDeliveryTime}}\)) | 0.3667 | hr/run | Mean delivery duration from operational deployments [39] |
| Max Delivery Time (\(T_{\text{MaxDeliveryTime}}\)) | 1 | hr/run | Maximum guaranteed delivery time [1], [37] |
| Avg Delivery Distance (\(T_{\text{AvgDeliveryDistance}}\)) | 4 | km/run | Mean delivery distance per SEC filings [29] |
| Max Delivery Distance (\(C_{\text{MaxDeliveryDistance}}\)) | 6 | km/run | Maximum operational delivery radius [38] |
| Charge Efficiency (\(\eta_{\text{charge}}\)) | 0.95 | unitless | Estimated charge efficiency [41] |
| Battery Roundtrip Efficiency (\(\eta_{\text{batteryroundtrip}}\)) | 0.95 | unitless | Estimated battery roundtrip efficiency [4] |
| Inverter Efficiency (\(\eta_{\text{inverter}}\)) | 0.99 | unitless | Estimated inverter efficiency [42] |
| Motor Efficiency (\(\eta_{\text{motor}}\)) | 0.82 | unitless | Estimated motor efficiency [56] |
Future Enhancements¶
| Enhancement | Description |
|---|---|
| Use the cost model as a reward signal for RL, enabling agents to directly optimize for profitability | |
| Model latency and bandwidth costs for cloud-based VLA policies | |
| Adjust revenue based on demand, distance, time of day | |
| Optimize across multiple robots (routing, charging) | |
| Factor in fines and liability costs from traffic violations |
References¶
| # | Reference |
|---|---|
| [1] | AGV Network. Starship Technologies FAQs. FAQ page, 2024. |
| [3] | Association for the Advancement of Automotive Medicine. Abbreviated Injury Scale: 2015 Revision. Chicago, IL, 6 edition, 2018. |
| [4] | Battery Design. Round-trip efficiency of lithium-ion batteries. Battery Design article, 2024. |
| [7] | City of Hot Springs. Frequently asked questions. Solid Waste FAQ, 2026. |
| [13] | Jared Worth. Maximizing humanoid robot longevity: the power of maintenance-free components. igus Toolbox article, 2024. |
| [15] | Lucas. The ROI of autonomous mobile robots in your DC. Lucasware article, 2020. |
| [21] | Multirotors.store. Segway outdoor delivery robot. Store Listing, 2025. |
| [22] | National Highway Traffic Safety Administration. The economic and societal impact of motor vehicle crashes, 2019. Report No. DOT HS 813 403, 2023. |
| [26] | Salary.com. Data collector salary information. Salary.com listing, 2025. |
| [27] | Salary.com. Robot operator salary information. Salary.com listing, 2025. |
| [28] | Segway Robotics. Segway E1 robot specifications (E1 delivery parameters). Specification PDF, 2025. |
| [29] | Serve Robotics Inc. Serve robotics inc. form 10-k for fiscal year 2023. U.S. Securities and Exchange Commission. Available at SEC EDGAR database. |
| [30] | Serve Robotics Inc. Serve robotics inc. form 10-q. U.S. Securities and Exchange Commission, 2024. Available at SEC EDGAR database. |
| [31] | Serve Robotics Inc. Serve robotics inc. form 424b4 prospectus. SEC EDGAR filing, 2024. |
| [36] | Standard Bots. How much do robots cost? 2026 price breakdown. Standard Bots article, 2026. |
| [37] | Starship Technologies. Starship technologies launches commercial rollout of autonomous delivery. Press release, 2024. |
| [38] | Starship Technologies. Starship dimensions. Dimensions spec page, 2024. |
| [39] | Starship Technologies. Starship technologies celebrates one year of robot deliveries in Finland. Press release, 2024. |
| [40] | Starship Technologies. Starship robot technical specifications. Starship Deliveries industry page, 2024. |
| [41] | STMicroelectronics. GaN-based Li-ion Battery Charger Demo. Seminar PDF, APEC 2023. |
| [42] | Texas Instruments. TIDA-010056: 48V 3-Phase Inverter. TI TIDA design page, 2024. |
| [43] | The Restaurant HQ. Food delivery statistics. Article, 2024. |
| [45] | Town of Wrentham, MA. Mailbox damage claims (news flash). Article, 2026. |
| [47] | University of South Carolina. Grubhub and Starship delivery. USC Dining Services page, 2024. |
| [48] | U.S. Energy Information Administration. Electric power report. EIA electricity data page, November 2025. |
| [51] | Wang, J.-S. Mais(05/08) injury probability curves as functions of delta v. NHTSA report PDF, May 2022. |
| [52] | We Are Bollards. How much does it cost to install a bollard? Blog post, 2024. |
| [54] | Wittenberg University. Wittenberg university damage charge schedule 2024-2025. Damage Charge Schedule PDF, 2024. |
| [56] | ZAU. M250W Hub Motor Specifications. ZAU motor spec page, 2024. |