CostNav

Ref: https://chatgpt.com/share/691ebb05-6d48-8004-ad22-2736e17a6be3

0. Common Environmental Parameters (U.S. Baseline)

(1) Order-Distance Distribution D_order

• For U.S. food delivery (all platforms), the median delivery distance ≈ 2.5 miles.

  • Source: Serve Robotics investor materials stating “U.S. food delivery median distance = 2.5 miles.” ([investors.serverobotics.com][1])

• Simulator defaults:

  • D_order_median ≈ 2.5 mi
  • D_order_mean ≈ 2.5–3.0 mi (with a long-tail including 4–5+ mi orders)

(2) Delivery-Time Distribution T_delivery (All Platforms)

Intouch Insight tested 600 orders in the U.S. (200 per app): ([foodondemand.com][2])

• All apps avg: T_delivery_all ≈ 33.4 min
• DoorDash: T_delivery_DD ≈ 26.4 min
• Uber Eats: T_delivery_UE ≈ 38.1 min

Approximate orders/hour:

• DoorDash: ≈ 2.27 orders/hour
• Uber Eats: ≈ 1.58 orders/hour
• Overall average: ≈ 1.8 orders/hour (Includes restaurant wait, parking, walking, etc.)

---

1. Human Couriers (DoorDash / Uber Eats) — Simulator Inputs

1-1. Basic Mobility Metrics

Distance d_h

• Based on U.S. median distance, d_h ≈ 2.5 mi is a natural baseline. ([investors.serverobotics.com][1])

• Simulator example:

  • DoorDash & Uber Eats shared baseline: d_h ~ N(μ=2.5 mi, σ≈1.0 mi) with tail orders >5 mi

Time T_h_total

• DoorDash: avg 26.4 min ([foodondemand.com][2])
• Uber Eats: avg 38.1 min ([foodondemand.com][2])
• Overall: 33.4 min

Simulator baseline example:

• T_h_total_DD_mean = 26.5 min
• T_h_total_UE_mean = 38.0 min
• Noise ±8–10 min to model restaurant delays & peak hours

---

1-2. Revenue & Cost Metrics (Platform + Courier Perspective)

(A) Driver Pay / Platform Payouts

Gridwise U.S. aggregation (2023Q1–2024Q2): ([Gridwise][3])

DoorDash

• Per-task pay: pay_h_task_DD ≈ $8.49
• Hourly earnings: ≈ $18.93/h

Uber Eats

• Per-task pay: pay_h_task_UE ≈ $10.00
• Hourly earnings: ≈ $24.68/h

Simulator baseline (platform payout):

• DoorDash: ≈ 8.5 USD/order
• Uber Eats: ≈ 10.0 USD/order (Tips are customer→driver, not a platform expense)

---

(B) Vehicle Cost per Mile

• IRS 2025 standard mileage rate: $0.70/mile ([IRS][4])

Simulator:

• c_vehicle_per_mile = 0.70 USD/mile

• Example vehicle cost per order:

  • If d_h_total ≈ 3.0 mi: C_vehicle_order = 0.70×3.0 ≈ 2.1 USD

---

1-3. Accident & Safety Metrics

(A) Accident Probability p_accident_h

Sentiance/industry estimate: ~1 accident per 143,000 deliveries. ([Food Delivery Apparel][5])

• p_accident_h_per_order ≈ 7.0 × 10⁻⁶
• DoorDash reports >99.99% of deliveries completed safely, consistent with the above. ([about.doordash.com][6])

(B) Cost per Accident

U.S. averages:

• Collision repair cost avg: ≈ $4,700+ ([Repairer Driven News][7])

• Liability claim averages:

  • Property damage: ≈ $6,551
  • Bodily injury: ≈ $26,501 ([III][8])

Simulator simplification:

• Minor accident: C_accident_minor ≈ 5,000 USD
• Full average incl. injury: C_accident_avg ≈ 20k–30k USD

Expected accident cost per order:

• Property-only: E[C_accident_prop] ≈ 0.046 USD/order
• Conservative full cost: E[C_accident_full] ≈ 0.23 USD/order

Practical simulator setting: 0.05–0.20 USD/order as tunable expected accident-cost parameter.

---

1-4. SLA & Promotions (Human Delivery)

Delivery Delay / SLA

Based on Intouch Insight: DoorDash is faster (26.4 min) than Uber Eats (38.1 min). ([foodondemand.com][2])

Informal industry norms:

• Delays >10–15 min over ETA → coupon/partial refund common.

Simulator example:

• L_h > 10 min → C_late_h ≈ $2
• L_h > 30 min → $5
• L_h > 60 min → $10

---

Promotions (Peak Pay / Surge)

DoorDash Peak Pay: +$1–$4/order, occasionally +$7.50. ([foodondemand.com][2])

Uber Eats Surge/Boost: Multipliers (1.3×, 1.5×) or flat fees (+2–5 USD).

Simulator:

• DoorDash: {0, 1, 2, 3, 4, 7.5} USD
• UberEats: multiplier 1.0–2.0 or flat 2–5 USD

---

2. Delivery Robots (Starship, Serve, Generic, Nuro) — Simulator Inputs

2-1. Generic Sidewalk Robots (Academic / Industry Source)

Typical small sidewalk robot (Berkeley CMR): ([California Management Review][9])

• Top speed: ≈ 4 mph
• Cruise: 3–4 mph
• Unit cost: ≈ 2,500–5,000 USD

Simulator:

• v_r_sidewalk = 3.5 mph
• C_robot_unit_generic = 4,000 USD

---

2-2. Starship Technologies (Sidewalk Robots)

Performance

• Top speed: ≈ 6 km/h (3.7 mph) ([California Management Review][9])
• Battery life: ≈ 18 hours
• Daily distance: ≈ 40 km/day ([TechCrunch][10])
• Payload: ≈ 20 lb ([Wikipedia][11])

Simulator:

• H_r_max_per_day_Starship = 18 h
• range_r_per_day_Starship = 40 km
• If each order ≈ 2–3 km → 12–15 orders/day

Capex / Autonomy

• Starship estimated unit cost:

  • 2018: ≈ 5,500 USD
  • Target: ≈ 2,250 USD ([Accio][12])

• Autonomy ≈ 99% ([starship.xyz][13])

Simulator:

• C_robot_unit_Starship ≈ 3,000–5,500 USD
• autonomy_ratio_Starship ≈ 0.99
• p_intervention_r_Starship ≈ 0.01

---

2-3. Serve Robotics (Gen2 / Gen3)

Distance Environment

• U.S. median food-delivery distance = 2.5 miles. ([investors.serverobotics.com][1])

• Serve-DoorDash sample:

  • Median Serve order distance ≈ 1.3 miles
  • Delivery time ≈ 18 min
  • Max speed ≈ 11 mph ([Investors.com summary][14])

Simulator:

• d_r_Serve ≈ 1.3 mi
• T_r_total_Serve ≈ 18 min
• v_r_cruise_Serve ≈ 5–8 mph

---

Operating Hours (Gen2 vs Gen3)

2024Q2 operational data:

• Daily supply hours: 385 h
• Active robots ≈ 48 → ~8 h/day per robot ([Restaurant Dive][15])

Gen3 improvements:

• Speed ×2
• Range ×2
• +6 hours operation per day ([Quiver Quantitative][16])

Simulator:

• Gen2: H_r_max/day ≈ 8 h
• Gen3: ≈ 14 h
• with doubled range & speed

---

Cost Structure

Investor presentation:

• Serve long-term goal: $1.00/order fully-loaded last-mile cost. ([investors.serverobotics.com][1])

Simulator:

• C_total_r_Serve_target ≈ 1.0 USD/order
• Early-stage scenarios: 1–3 USD/order

Reliability

Serve reports operational efficiency ~99.8% (vs human ~98%). ([Barron’s][17])

Simulator:

• p_failure_r_Serve ≈ 0.002
• p_failure_h ≈ 0.02

---

2-4. Nuro R2 (Road-Driving Robots)

• Road speed: ≈ 25 mph (R3 → 45 mph) ([SlashGear][18])
• Battery: ≈ 31 kWh, city range tens to >100 km

Simulator:

• v_r_road_cruise ≈ 20 mph
• range_r_road ≈ 100–150 km/day
• C_robot_unit_road ≳ 50,000 USD (treated as a scenario parameter)

---

3. Humans vs Robots — Shared KPI Inputs

3-1. Total Cost per Order C_total

Human (DoorDash/Uber Eats)

• Driver payout: ≈ 8.5–10 USD ([Gridwise][3])
• Vehicle cost: ≈ 2.1 USD (3 mi × $0.70) ([IRS][4])
• Accidents: ≈ 0.05–0.20 USD/order ([Food Delivery Apparel][5])

Baseline:

C_total_h_DD ≈ 8.5 + 2.1 + 0.1 = 10.7 USD/order
C_total_h_UE ≈ 10.0 + 2.1 + 0.1 = 12.2 USD/order

Robots (Serve / Starship)

• Serve goal: ≈ 1.0 USD/order ([investors.serverobotics.com][1])

Starship estimation:

• Depreciation: ≈ 0.3–0.8 USD/order
• Energy: ≈ 0.01–0.03 USD/order ([TechCrunch][10])
• Maintenance + teleop → target 1–3 USD/order ([California Management Review][9])

Baseline:

C_total_r_Starship ≈ 1.5–2.0 USD/order
C_total_r_Serve   ≈ 1.0–2.0 USD/order (target: 1.0)

---

3-2. Productivity / Utilization

Human Drivers

• DoorDash: ≈ 2.27 orders/h ([foodondemand.com][2])
• Uber Eats: ≈ 1.58 orders/h
• Full 8h shift: ≈ 12–18 orders/day

Robots

• Starship: 18h/day, 40 km/day → ≈ 12–15 orders/day ([TechCrunch][10])
• Serve Gen2: ≈ 8h/day
• Serve Gen3: ≈ 14h/day ([Restaurant Dive][15])

---

Usage in Simulator

Using this style — listing (DoorDash: X, Uber Eats: Y, Starship: Z, Serve: W) next to each variable — allows consistent comparison across scenarios:

• Baseline
• Human-only
• Robot-only
• Mixed fleets

If you want, I can also produce:

• YAML/JSON configuration templates
• Python dataclasses for these parameters
• Auto-calibration functions for cost curves

---

All original references remain unchanged:

[1] https://investors.serverobotics.com/static-files/0138a856-bd69-4564-9f8d-0534f9cebb73 [2] https://foodondemand.com/09262024/doordash-tops-intouch-insight-report-on-delivery-performance/ [3] https://gridwise.io/blog/delivery/uber-eats-vs-doordash-pay-how-much-are-drivers-earning/ [4] https://www.irs.gov/tax-professionals/standard-mileage-rates [5] https://fooddeliveryapparel.com/blogs/news/the-hidden-dangers-when-youre-on-the-road-the-reality-of-being-a-food-delivery-driver [6] https://about.doordash.com/en-us/news/how-were-making-dashing-even-safer [7] https://www.repairerdrivennews.com/2024/12/20/ccc-top-2024-trends-perfect-storm-of-softening-economy-increasing-repair-costs-and-aging-vehicle-pool/ [8] https://www.iii.org/fact-statistic/facts-statistics-auto-insurance [9] https://cmr.berkeley.edu/2022/04/self-driving-robots-a-revolution-in-the-local-delivery/ [10] https://techcrunch.com/2022/01/25/starship-technologies-picks-up-e50m-from-the-eus-investment-arm-to-expand-its-fleet-of-autonomous-delivery-robots/ [11] https://en.wikipedia.org/wiki/Starship_Technologies [12] https://www.accio.com/plp/wheeled-delivery-robots [13] https://www.starship.xyz/ [14] https://www.investors.com/news/technology/serve-robotics-doordash-partnership/ [15] https://www.restaurantdive.com/news/serve-robotics-debuts-los-angeles-faster-larger-robots-uber-eats-deal/729973/ [16] https://www.quiverquant.com/news/Serve%2BRobotics%2BSecures%2BTop%2BSpot%2Bin%2BFast%2BCompany%27s%2B%22Next%2BBig%2BThings%2Bin%2BTech%22%2Bfor%2BThird-Generation%2BAutonomous%2BDelivery%2BRobot [17] https://www.barrons.com/articles/serve-ai-robot-uber-stock-results-beb6ba3d [18] https://www.slashgear.com/nuro-r2-self-driving-delivery-pod-hits-us-autonomous-vehicle-milestone-06608901/