How do I choose between different mobile robot autonomy stacks for campus delivery and yard logistics?

Choosing between mobile robot autonomy stacks for campus delivery and yard logistics is mostly about matching the stack to the operating environment, safety requirements, fleet size, and how much you want to customize.

1) Start with the mission profile

Ask these first:

• Environment: sidewalks, roads, parking lots, indoor/outdoor transitions, unstructured yards, gravel, mud, snow?
• Speed: walking pace vs low-speed industrial vs mixed traffic?
• Navigation complexity: GPS available? Tree cover? Tight aisles? Dynamic pedestrians/vehicles?
• Payload and docking: light parcels vs totes vs carts; need lift/drop, locker integration, charging dock?
• Fleet scale: 1–3 pilots or 50+ robots?
• Regulatory/safety needs: public right-of-way, pedestrian interaction, functional safety, site certification?
• Customization: do you need to modify perception/planning deeply, or mostly configure and integrate?

2) Common stack “types”

A. Commercial turnkey autonomy stack

Best when you want fast deployment and lower engineering burden.

Good for:

• Campus delivery pilots
• Yard logistics in controlled sites
• Teams without a large robotics software group

Pros:

• Faster time to market
• Usually includes localization, obstacle avoidance, fleet management hooks, remote ops tools
• Vendor support and field experience

Cons:

• Less control over algorithms
• Can be expensive
• Integration/customization constraints

B. Open-source / modular autonomy stack

Best when you want flexibility and ownership.

Good for:

• Research-heavy or highly customized robots
• Teams with strong robotics/software engineering
• Unique hardware or unusual environments

Pros:

• Highly customizable
• Lower software licensing cost
• Easier to inspect and modify components

Cons:

• More integration work
• You own reliability, testing, and maintenance
• Often needs substantial systems engineering for deployment

C. Hybrid stack

A lot of successful deployments use:

• Open-source middleware and navigation
• Commercial perception/fleet tools
• Custom mission logic and remote ops

Best when:

• You want to avoid lock-in but still need production-grade pieces
• You’re scaling from pilot to operations

3) Key evaluation criteria

Use these to compare stacks side by side.

Localization and mapping

• Works with GNSS / RTK, LiDAR, vision, or sensor fusion
• Handles GPS-denied areas like courtyards, under trees, near buildings
• Supports map updates and re-localization after environment changes

Navigation in dynamic environments

• Pedestrian avoidance
• Vehicle and forklift interaction
• Narrow paths and temporary obstacles
• Behavior at curb cuts, gates, ramps, loading zones

Terrain handling

For yard logistics, this is huge:

• Uneven ground, potholes, gravel, grass, mud
• Slip detection and recovery
• Chassis and suspension requirements
• Ability to use terrain-aware planning if needed

Safety and compliance

• Emergency stop behavior
• Speed limiting and geofencing
• Obstacle detection coverage
• Audit logs and incident replay
• Support for site risk assessments and operational safety cases

Fleet operations

• Mission dispatching
• Charging management
• Remote teleoperation or intervention
• Health monitoring and alerting
• OTA updates and version control

Hardware compatibility

• Sensor support: LiDAR, radar, cameras, IMU, GNSS
• Compute platform support
• Drive-by-wire or motor controller interfaces
• Support for your robot form factor and payload

Reliability and maintainability

• Recovery behaviors when sensors fail
• Watchdogs and degraded modes
• Diagnostics and observability
• Update strategy without bricking fleets

4) Campus delivery vs yard logistics: what matters most

Campus delivery

Prioritize:

• Pedestrian-aware navigation
• Sidewalk and curb handling
• Weather robustness
• Regulations and public-space safety
• Quiet operation and good human interaction

Stack preference:

• Strong perception stack
• Good map management
• Mature remote operations tools
• Vendor support can be very helpful here

Yard logistics

Prioritize:

• Rough terrain and localization robustness
• GNSS/RTK plus local sensing
• Simple, reliable mission execution
• Integration with yard workflows, gates, trailers, and loading equipment

Stack preference:

• Strong low-level vehicle control
• Sensor fusion for mixed conditions
• Flexible autonomy architecture
• Often more emphasis on industrial reliability than human-interaction behaviors

5) A practical decision rubric

Rate each candidate stack from 1–5 on:

• Time to deploy
• Localization robustness
• Dynamic obstacle handling
• Terrain capability
• Safety tooling
• Fleet ops maturity
• Hardware compatibility
• Customization freedom
• Total cost of ownership
• Vendor/community support

Then weight them:

• For campus delivery: safety, human interaction, remote ops, deployment speed
• For yard logistics: terrain, robustness, integration, uptime, cost

6) Red flags

Be cautious if a stack:

• Works only in ideal conditions
• Has weak failure recovery
• Needs heavy hand-tuning for every site
• Lacks good logs/telemetry
• Doesn’t support your sensor set
• Has no clear answer for public-space safety or site certification
• Cannot scale beyond a small pilot

7) Recommended approach

If you’re early:

1. Pick 2–3 candidate stacks
2. Test on your worst-case route, not a demo route
3. Measure:
   • success rate
   • intervention rate
   • mean time between failures
   • map maintenance burden
   • operator workload
4. Run a small pilot in real conditions
5. Choose the stack that minimizes operational pain, not just demo performance

8) Rule of thumb

• Choose commercial turnkey if you need speed, support, and lower integration risk.
• Choose open-source/modular if you need control, customization, and long-term flexibility.
• Choose hybrid if you want a production path without giving up all control.

If you want, I can also give you:

1. a comparison matrix of common autonomy stacks, or
2. a decision tree for campus delivery vs yard logistics.