i had search over google and found out more or less some sensor that i could use like sonar, gyroscope, accelerometer,magnetometer and gps and 2d lidar (which is less expensive, around $300) i have cost budget so i couldnt go for those expensive devices.

im new to robotic. thus i want to ask for some recommendation. i want to build a robot which satisfied these following feature enter image description here

  1. works on outdoor. look at the image above with some description here: blue line is the border. red line is the path. black line is the building.
  2. when the robot at point A, and i input point B. the robot should be able to find a path from point A to B
  3. if point x is blocked, it should be able to find a way to point B via point Y
  4. ground is not flat. thus there might be need sense some higher ground level to avoid collision and recalculate the position of the robot
  5. works on day and night

therefore, my question is: what is the best device/sensor i should get to satisfy my features above. or probably any technique i could use? Please kindly share with me. maybe there is something i miss out on my research

Gps is good but the error is around 6m.. that is too large fusion of gyroscope, accelerometer, magnetometer to get a correct position and direction is kinda difficult(oblu maybe)

founding: i found 2 device that could help me with that

  1. ZED stereo camera which works on 0.5m up to 20m.($499)

  2. Intel RealSense 435i which works 0.1m up to 10m with IR.($200) both are not working so good in dark area even with IR. but i think a lighting could be installed to help with that

any other device recommendation ?

  • $\begingroup$ can you use any kind of odometry and a camera? $\endgroup$
    – galtor
    Commented Feb 11, 2019 at 20:56
  • $\begingroup$ Welcome to Robotics zhang yu, but I'm afraid that Unbounded Design Questions are off-topic because there are many ways to solve any given design problem. We prefer practical, answerable questions based on actual problems that you face, so questions which ask for a list of approaches or a subjective recommendation on a method (for how to build something, how to accomplish something, what something is capable of, etc.) are off-topic. Please take a look at How to Ask & tour for more information on how stack exchange works. $\endgroup$
    – Chuck
    Commented Feb 12, 2019 at 12:59
  • $\begingroup$ I would suggest you look at localization algorithms and see what packages are typical for those solutions, or simultaneous localization and mapping (SLAM) if you want the robot to build the map by itself. You said you're intending to give a destination, which would imply you and the robot already share a common coordinate frame. You mention at the top that \$300 is to expensive and then go on to mention a \$500 sensor as an alternative. Once you know what kind of sensor you want and can fix a price budget, the rest of the task is looking around until you find the thing that meets your needs. $\endgroup$
    – Chuck
    Commented Feb 12, 2019 at 13:05
  • $\begingroup$ Thanks @Chuck the main issue here were deciding the hardware to use. what i mention above $300 was the for the lidar. while on lidar alone, it couldnt satisfy all my needs. that is where i count as expensive. however i had decided to go on Intel RealSense 435i. which is satisfy all my requirements. $\endgroup$
    – zhang yu
    Commented Feb 18, 2019 at 3:08

2 Answers 2


This is a broad topic you've opened up here. My contribution:

  1. It all comes down of the robot specs. For example, is it a wheeled robot or a legged.? What size of terrain shall it explore? in the scale of meters, centimeters, tens of meters? etc... I think, you should write down all questions you could think of, that will help you specify your sensor options.

  2. I think you need a redundant sensor system. For adequate precision it is inevitable. Multiple sensors that are inaccurate on their own, but combined they give accurate results. Also your sensor data and control data shall be piped through some sort of filtering algorithm to increase performance. It will gain complexity, but that is your tradeoff for accuracy.

  3. You could for example have a nice IMU sensor, it provides 3D accelerometer and compass data. That you can use to determine your relative position, speed, acceleration.

  4. Are you using a wheeled robot? You could use the wheel rotation to determine the relative path your robot has achieved. It may be really inaccurate, since the given terrain, wheel skidding etc.

  5. If you have the two above you may as well can apply an IR or ultrasound sensor, to increase precision.

  6. You still can use a cheap mono camera, to estimate speed and direction.

  7. Also important, how you utilize your sensor data. Building a nice visual map based on the sensor data could help a lot in navigation. AI that can do some extra analysis on the shapes on the map.

Probably, most of the things above were also known for you, what I tried to show here is, that if you want to achieve high or just adequate performance, You can't get away with a simple mighty sensor.

As a concrete example: my master work was a multilegged robot with a single sweeping ultrasound sensor. It had an algorithm, that has built up a map based on the sensor input and run the analysis on it. It actually have solved relatively complex static and dynamic maze problems using the real robot hardware. But, it has flawed accuracy, caused by the monosensory and error in calculated position, that obviously can't reflect measured position, in case of a legged robot, without additional sensors. I hope I can at least give you some ideas of thinking.

  • $\begingroup$ Thanks, I might be not clear enough with my question. and yes, we had considered several factors. and this is why adometry cant be used here. so i had decided to go on RealSense D435i thanks alot guys $\endgroup$
    – zhang yu
    Commented Feb 18, 2019 at 3:35

The question is about the hardware of a robot. What is not asked is how to program a robot by software, this would be an Artificial Intelligence question. I belief, that software is more interesting but to answer the question we have to stay at the hardware side. Hardware and mechanics were invented in the 19th century during the industrial revolution. Toys from that area are sold under the brand name Meccano. If somebody likes to build physical robots or is interested in hardware devices, this is a reference. According to the website, Meccano robots have an infrared sensor as a standard sensor for detecting obstacles. This device can be used for indoor and outdoor problems as well. A dedicated outdoor infrared sensor is often named as laser/IR sensor.

So my recommendation is to focus only on one sensor type which is infrared and use this hardware for obstacle detection, navigation and pathplanning at the same time.

  • $\begingroup$ Thanks i had decided to go on RealSense 435i $\endgroup$
    – zhang yu
    Commented Feb 18, 2019 at 3:33

Not the answer you're looking for? Browse other questions tagged or ask your own question.