We've been building low-speed autonomous delivery robots for about two years, mainly serving industrial parks, university campuses, and factory complexes. Our early units used a 2D LiDAR + ultrasonic sensor combo that worked okay but had real limitations. We've since moved to full 3D perception, and the rooftop unit of choice is the Tantu M360 — tested head-to-head against a Livox MID-360. Here's what six months taught us.

What delivery robots actually deal with

Our delivery robots are box-style vehicles, roughly 1.2m long, 0.8m wide, and 0.9m tall, carrying up to 50kg at speeds of 0.5–1.5 m/s. They deliver parcels, meals, and documents.

Compared to AGVs and forklifts, delivery robots operate in a very different environment:

Why we ditched 2D + ultrasonic for full 3D

Our original setup: 2D single-line LiDAR for horizontal obstacle avoidance, ultrasonic sensors for close-range backup. It worked, sort of.

The big problem: 2D LiDAR is blind to height. One day a robot was driving down a park road, the 2D LiDAR showed a clear path ahead, and it drove straight into a tree branch extending over the lane. The branch sat just above the scanning plane. The ultrasonic sensors, mounted low on the front bumper, couldn't reach it either.

After that, we went 3D. Tree branches, overhead signs, and the rear ends of parked SUVs all became visible. We dropped the ultrasonic sensors and simplified the sensor stack.

M360 vs MID-360 on delivery robots

Both units were mounted on the roof, slightly forward of center, at roughly 0.9m height with a 10° downward tilt.

Blind zone: 5cm vs 10cm

Our robots have about 15cm ground clearance. Common low obstacles — curb edges, speed bumps, fallen shared-bike pedals, pets — typically sit in the 5–15cm range.

With a 10cm blind zone, obstacles between 5–10cm are almost invisible. On a 200m test stretch of park road, the M360 avoided 8 low obstacles; the MID-360 avoided 6. The two misses: a fallen bike pedal (~6cm) and a brick fragment at a curb edge (~4cm).

The 6cm pedal won't necessarily damage the robot, but at 15cm ground clearance it could snag the undercarriage. The 4cm brick probably wouldn't cause damage, but the jolt would destabilize the cargo inside.

Vertical FOV: 70° vs 59° — a significant gap here

At just 0.9m mounting height with a 10° downward tilt, the LiDAR needs to cover both the ground surface (curbs, steps, low obstacles) and the space above (tree branches, signs, vehicle rears).

The M360's 70° vertical FOV covers from roughly 0.3m above the ground up to about 1.5m overhead. Nearly complete volumetric coverage around the robot.

The MID-360's 59° loses 11°, capping out at around 1.2m overhead. Taller obstacles — SUV rears, tree branches — may sit at the edge of the scanning volume with sparse point coverage.

On tree-lined campus roads, the difference is noticeable. The M360 consistently produces cleaner point clouds of overhead obstacles.

Weight and dimensions

265g vs 408g — the MID-360 is 143g lighter. Our robots weigh about 80kg, so either LiDAR is a small fraction. But for ultra-compact last-mile robots where every gram matters, the MID-360 has an edge.

On dimensions, the M360 (78×78×81mm) is noticeably larger than the MID-360 (65×65×60mm). In tight rooftop mounting scenarios, the MID-360 fits more easily.

Power consumption

Delivery robots typically run 24V or 48V battery packs of 30–50Ah. Total vehicle power is around 100–200W (drive motor + compute + sensors).

The 2W difference (4.5W vs 6.5W) over a 12-hour shift is 24Wh — under 1% of a 48V/50Ah pack. Not a deciding factor on its own, but when you're running LiDAR + cameras + ultrasonics, the savings add up.

Input voltage — cold mornings matter

Our robots feed sensors directly from 24V battery packs. The M360's 12–32V range is a comfortable fit; the MID-360's 9–27V also works at nominal voltage.

But in cold weather, battery voltage drops. At sub-zero temperatures, a 24V lead-acid battery at full charge might read only 25V, dropping to 23V at half charge. The MID-360's 27V upper limit is fine at these voltages, but the margin is noticeably thinner than the M360's 32V.

We had one incident where a robot sat outdoors overnight in winter. The MID-360 threw a low-voltage error on startup and didn't initialize until the battery was recharged. The M360 started without issues.

Dual return — rain on tree-lined roads

Delivery robots spend most of their time outdoors, so dual return matters more than it does for indoor-only platforms.

Our campus has a tree-lined avenue. In rain, water dripping from the canopy creates a storm of noise in single-return mode — especially under dense foliage. Dual return captures the forward echo (off raindrops) and separates it from the return echo (off actual surfaces). Tree trunks, speed bumps, and parked vehicles stay clearly visible.

We do reduce speed or suspend operations in heavy rain. Dual return is primarily a safety net, not a daily necessity. But it has helped maintain safe operation during light rain and drizzle when full suspension would be an overreaction.

Non-repetitive scanning at intersections

Delivery robots move slowly and stop frequently at intersections to wait for pedestrians or traffic. Non-repetitive scanning builds point cloud density during these stops.

Intersections are the most complex scenario: cross-traffic, pedestrians, curb edges, and manhole covers all need simultaneous attention. Both LiDARs support non-repetitive scanning. After 2–3 seconds stationary, the point cloud density around the robot improves noticeably, and small obstacle detection becomes more reliable.

IP67 and operating temperature

Both units carry IP67, which is essential for outdoor robots exposed to rain, dust, and mud.

One detail worth noting: delivery robot LiDARs sit fully exposed on the roof with no vehicle body shielding. In direct summer sun with the sensor's own heat generation, surface temperatures can exceed 60°C. The M360 is rated to +60°C; the MID-360 to +55°C.

The MID-360's upper limit is 5°C lower. We haven't experienced a heat shutdown in our climate, but teams deploying in southern China or tropical regions should factor this in.

Our current setup

All six park delivery robots run the M360 (two are M360-D dual-return variants).

Sensor stack:

We chose the M360 over the MID-360 primarily for three reasons: the 70° FOV provides more complete volumetric coverage at our low mounting height, the 12–32V input range handles cold-start voltage drops better, and dual return adds a rain safety margin.

For ultra-compact last-mile robots where weight and size are critical, the MID-360 is a perfectly valid choice.

Selection advice

Key specs to prioritize for delivery robot LiDAR:

Both the M360 and MID-360 can handle delivery robot requirements. The M360 has an edge in FOV and power input range; the MID-360 is smaller and lighter. The right call depends on your robot's size and operating environment.

*Based on our field testing. MID-360 specs from Livox's official website; M360 specs from Tantu Smart Mobility. Results may vary by vehicle type and environment.*