What is the principle of drone automatic obstacle avoidance?

the principle of drone automatic obstacle avoidance

The operation of the UAV automatic obstacle avoidance system can be divided into three stages. The first is to perceive obstacles; the second is to bypass obstacles; the third is path search, that is, the process of drones from discovering obstacles to automatically avoiding obstacles to automatically planning flight routes.

In the first stage, the UAV senses the obstacle, quickly recognizes it and hovers over it, waiting for the next instruction.

In the second stage, the UAV accurately perceives the specific outline of the obstacle through the sensor, and then bypasses the obstacle autonomously.

In the third stage, the UAV uses the acquired environmental information to automatically plan the flight route using algorithms, so as to realize the function of automatic automatic obstacle avoidance.

To put it simply, "binocular distance measurement" is a routine operation, but it contains N sets of algebraic equations. In the early driverless cars, this logic operation method was also used, but due to the huge processing capacity , the car is equipped with multiple top-level computers for collaborative calculations. Can make accurate judgments in a short period of time.

On UAVs, general UAV flight control processors cannot achieve such a computing speed, and manufacturers that have been developing mobile phone processors (such as: Intel, Qualcomm), relying on the low energy consumption and high performance features that become their entry point, like using Intel processing, Yuneec Typhoon H, DJI Matrice M100 GUIDANCE accessories, and ZERO XPLORER with Qualcomm Snapdragon processor

2. Tencent Ying, DJI Phantom 4 and other products.

Because of the powerful computing power, the 3D environment model can be drawn in real time, and more intelligent optimization functions can be realized, just like the "pointing and flying" function realized on the DJI Phantom 4, which is also a 3D model built by relying on the forward-looking obstacle sensor To achieve this operation, also in follow mode, no one
Robots will be more accurate when following tasks within visual range. In the process of automatic return, automatic obstacle avoidance is also relatively safer. There is finally a more reliable solution to the things that used to hit walls, trees, and mountains.

However, due to the large amount of calculation, the voltage will increase correspondingly during the calculation, the energy consumption will also increase immediately, and the battery life will also decrease accordingly. For each additional obstacle avoidance direction, the calculation amount will double, but it can draw a clearer 3D Model. Among the released drone products, the DJI Matrice M100 GUIDANCE accessories used by positioning developers can realize the automatic obstacle avoidance function in 5 directions, while the Typhoon H, Phantom 4, Tencent Ying and other drones targeting consumer products , the automatic obstacle avoidance function only supports the visual detection of the front.

The advantage of only forward obstacle avoidance is to ensure the battery life and realize the automatic obstacle avoidance function to the greatest extent. After all, the drone mainly flies forward, not to say that the drone will never hit the wall, and it also has its limitations. When in the smart following mode, the drone follows the person in front, the person walks forward, the drone will move backwards, the drone cannot recognize the obstacles behind it, and it is very likely that the drone will hit the wall .

Defects of the three major automatic obstacle avoidance drones

In addition, the visual impairment sensor used on drones has a certain range of use. DJI's Phantom 4 supports a range of 0.7 - 15 meters, and the lighting conditions must be above 15 lux, and it is invalid in low light. The Zero Degree Xplorer 2 is somewhat different, using the principle of laser radar TOF ranging, The "mushroom head" on the top of the head emits lidar, which scans 360° at a rate of 50 times per second to draw a 3D model. obstacle distance. And Yuneec Typhoon H uses Intel's Real Sense technology, Monocular distance measurement + structured light, the effective distance measurement can reach 10 meters, and the left infrared sensor and right infrared sensor can be used to track the position of this beam of light.