What makes outdoor handling more difficult than indoor handling?

We have worked on indoor AGV projects for several years, and last year some customers asked if we could make an outdoor version. Initially, we thought it should be similar, just moving the indoor solution outdoors. However, after actually doing it, we found that it's completely different.

The Essential Difference Between Indoor and Outdoor

Indoor environments are controllable – the ground is flat, lighting is stable, temperature is constant, and there is no rain or wind. The radar can operate in such an environment almost blindfolded.

Outdoors is different. The ground may be asphalt, gravel, mud, or grass; lighting changes from morning to night; after rain, the ground reflects light; in winter, the road surface may freeze; in summer, the sun can directly shine on the radar housing, raising the temperature to 50~60°C.

When moving an indoor solution to the outdoors, the first problem that usually arises is positioning. Indoors, SLAM is done with LiDAR, and features such as walls, shelves, and columns are stable and unchanging, making navigation very stable after the map is built. Outdoors, however? The grass on the lawn grows every day, tree shadows move with the sun, and parked cars may have left and been replaced by another. The static map quickly becomes outdated.

Therefore, when we developed the outdoor handling solution, the selection criteria for several key sensors were completely different from those for indoor solutions.

Three key capabilities of the M360 in outdoor scenarios

We ultimately chose the M360, not because of any particular parameter standing out, but because the combination of several capabilities matched the needs of the outdoor environment.

The first is dual echo.This is not very useful indoors, but it is very useful outdoors.

The most common interference outdoors is raindrops and mist. Handling goods in rainy weather is a necessity—the customer will not stop shipping just because of the rain. When single echo radar encounters raindrops, it will generate a large number of false echoes, and the point cloud is full of noise, making it difficult for the algorithm to distinguish between real obstacles and raindrops. This is like driving in a heavy rainstorm with a broken windshield wiper, and the windscreen is full of water, making it impossible to see what's ahead.

The dual-echo mode of the M360 can partially solve this problem. Raindrops are very small, and the first echo hits the raindrop, while the second echo passes through the raindrop and hits the real object behind it. At the algorithm level, by comparing the distance difference between the first and second echoes, the noise points from raindrops can be filtered out.

The results are not perfect—the noise is still significant during heavy rain, but under moderate rain and light rain, it can work normally. We have tested that under light rain (rainfall < 5mm/h), the point cloud quality with dual echo mode is almost the same as on sunny days; under moderate rain (5~15mm/h), there are a few noise points but they do not affect navigation; under heavy rain (>15mm/h), it is recommended to slow down or pause operations.

The second is working temperature.The operating temperature range of the M360 is -10°C to +60°C, which is sufficient for most outdoor scenarios.

Why is temperature important? The LiDAR internally has laser emitters and photoelectric detectors, which are sensitive to temperature. High temperatures can reduce the emission power of the laser, weakening the ranging capability; low temperatures can slow down the response speed of electronic devices, potentially reducing the frame rate of the point cloud.

In the northern winter, it is common for outdoor temperatures to drop below -10°C. If the radar's minimum operating temperature is only -20°C, it may seem like there is a margin, but in reality, during the cold start-up phase (when the radar is taken from indoors to outdoors), the rate of temperature change is very fast, and the internal components may not have had enough time to adapt to the environmental temperature. The -10°C lower limit of the M360 is generally sufficient for daily use – after all, outdoor handling usually does not operate under extreme low temperatures, and customers will also provide insulation for the machine.

In the southern summer, it is even more important to pay attention. Under direct sunlight, the temperature of the radar housing may be 20~30°C higher than the ambient temperature. If the ambient temperature is 35°C, the housing temperature may reach 55~65°C. The +60°C upper limit of the M360 is quite tight here. Our suggestion to customers is to add a sunshade to the radar, which can reduce the housing temperature by about 10°C.

The third is IP67.Outdoor transportation is almost all-weather operation, with IP67 as the standard. Not only is it rainproof, but also dustproof—the dust in places like construction sites, mines, and docks is much greater than indoors. The complete dustproof capability of IP67 is much more reliable in this environment than IP54/IP65.

We previously tested an IP54 radar, which was placed outside the workshop for two months. Upon inspection, we found a layer of fine dust on the lens. After cleaning it off, we discovered that the dust had penetrated the gaps of the optical components, making it difficult to clean thoroughly. The M360 with an IP67 rating, used for the same duration, was found to be completely clean inside upon inspection.

The use of a 5cm blind zone in outdoor settings

The obstacles in outdoor scenarios are different from those indoors. Indoor obstacles are mainly shelves, columns, and walls, which are regular in shape and have clear boundaries. Outdoor obstacles are more varied – stones, steps, curbs, fire hydrants, low hedges, and ground manhole covers.

These obstacles have a common characteristic: small size and low height. Curbs are generally only 10~15cm high, fire hydrants are 50~80cm tall, and manhole covers may only protrude 2~3cm.

The 5cm blind zone of the M360 has a significant advantage over the 10cm blind zone of the Mid-360 in this environment. With a curb stone 15cm high, a radar with a 10cm blind zone can detect 5cm earlier, while a radar with a 5cm blind zone can detect 10cm earlier. Don't underestimate this 5cm lead time – at a speed of 1m/s, 5cm means an additional 50ms of reaction time for the robot.

In outdoor handling scenarios, the speed of the robot is generally 0.5~1.5m/s, which is not very fast, but the coefficient of friction on outdoor ground is lower than that indoors (especially on asphalt or tile surfaces after rain), resulting in a longer braking distance. An extra 50ms of reaction time may be the difference between hitting a curb and avoiding it.

Power supply and actual wiring

The power supply solution for outdoor handling robots is more complex than indoor ones. Indoor AGVs usually use 24V lithium batteries for stable power supply. Outdoor handling robots, due to the need to handle heavier loads and longer travel distances, often use 48V or even higher battery solutions.

The power supply range of the M360 is 12~32V, and a battery with 48V cannot be connected directly; a voltage reduction module is required. The power supply range of the Mid-360 is narrower at 9~27V, and it cannot even accommodate a fully charged 8S lithium battery at 33.6V, let alone 48V.

Therefore, both radar systems require voltage reduction under the 48V scheme, with no fundamental difference. However, the 32V upper limit of M360 is wider than the 27V of Mid-360, allowing direct connection at 24V and 33.6V common battery solutions, thus saving the voltage reduction module and reducing one potential point of failure.

GPS and Inertial Measurement Unit (IMU) integration

Outdoor handling usually does not rely solely on LiDAR. GPS (or Beidou) provides global positioning, inertial navigation (IMU) provides short-term attitude estimation, and LiDAR provides local fine perception and obstacle avoidance.

The M360 is equipped with a 6-axis IMU, which can directly output acceleration and angular velocity data. This data can be fused with external GPS or serve as a supplementary data source for the IMU system.

In areas where GPS signals are obstructed (such as next to buildings, under bridges, or inside tunnels), the M360's SLAM can take over positioning to keep the robot navigating. After the GPS signal is restored, it switches back to GPS positioning. This seamless switching is common in outdoor scenarios and is also one of the reasons why outdoor solutions are more complex than indoor solutions.

A real comparison

A customer previously used the Mid-360 for an indoor handling project, with good results. When making the outdoor version, they initially directly used the Mid-360, but encountered two problems:

First, the point cloud quality is poor during rainy days. Their warehouse is outdoors, and sometimes it rains during loading and unloading, causing Mid-360 to have a large number of noise points in the rainy point cloud, frequent misjudgments by the obstacle avoidance algorithm, and the robot stopping and not moving after a few steps.

Second, the shell temperature is high after being exposed to the sun. After working outdoors for two or three hours in the midday summer sun, the radar shell feels hot to the touch, and point cloud data occasionally gets lost. It was found that the working temperature limit of Mid-360 is +55°C, and the laser will automatically reduce power to protect itself when the shell temperature exceeds this value.

After switching to the M360, the point cloud quality improved significantly during rainy days with dual echo mode. On light rain days, it barely affects the operation. In terms of temperature, the upper limit of +60°C offers an additional 5°C margin over Mid-360, and with the sunshade in place, there has been no power reduction.

However, the M360 is not perfect either – its ranging capability of 50m is shorter than the Mid-360's 70m. In outdoor scenarios, 50m of ranging is more than enough. For material handling robots, 50m is absolutely sufficient. The safe braking distance of material handling robots generally does not exceed 3m (at a speed of 1.5m/s), and the 50m detection distance has enough margin. The difference between 70m and 50m is imperceptible in handling scenarios.

The above is based on actual project testing; the M360 parameters are to be taken as the official data from TanTu Zhixing.