LeddarTech recently had the opportunity to collaborate with youtuber and drone aficionado Rai Gohalwar about the LeddarOne single channel sensor module and its benefits for rangefinding applications. We also took the time to have a little chat with Rai to learn more about where his tinkering skills come from, and how he foresees the future of the drone industry and next-generation collision avoidance capabilities.
But first, watch Rai’s review of the LeddarOne sensor module:
Hi Rai! Could you please first tell us how did you become interested in drones and related technologies?
Hi! As a child, I was always tinkering with batteries and wires. I distinctly remember in grade 5 taking apart small handheld DC fans and putting its guts in a Coca-Cola bottle to make a boat. My dad bought me a soldering iron and after burning myself a few times I learned how to use it properly. It wasn’t until grade 7 when my uncle purchased me a RC toy plane. It flew for 5 seconds and crashed in a tree. Later, my dad bought me another RC plane which wouldn’t even take off. As any child, I had no patience so I gave up on this hobby. I still had two broken RC airplanes, so being me I ripped them open and started exploring what was inside them. It all seemed so complex but fun, so I decided to build a foam plane and put the electronics I had laying around. After another two years of unsuccessful take-offs and crashes, I finally had a plane that flew. From that point on I was hooked. I had grown up watching Dave at RCPowers on Youtube. He pioneered the RC community on Youtube and then Flitetest followed, giving me even more inspiration to push my boundaries. I graduated to hover drones in grade 12. I build my first tri-copter which flew for 2 1/2 minutes. From that point on, there was no stopping me. I got a job in Ottawa to build a drone to fly over crops. There I was exposed to the industrial applications of drones.
Various technologies come into play when it comes to drone spatial awareness. What can you tell us about these technologies (namely, radar, LiDAR, camera, ultrasound, etc.) and what are each technology’s strengths and weaknesses?
Lets start from the basics. One of the greatest booster in the drone technology were the MEMS sensors. They were suddenly cheap, small and robust to be used on drones. Very soon many people realized this and designed flight controllers to be put on drones. The original ArduinoPilot was a cluster of sensors connected to an Arduino to control a drone. Even today, companies strive to make their sensors more robust and accurate. This is a vital factor in great UAV performance. It always amazes me to think how much sensors on top of a basic stack can affect a UAV’s performance.
Lets talk about cameras. Visual cameras are used to record flights and FPV fly for a long time. The real magic happens when the camera is facing down to the ground. Using the visual data captured from the CMOS sensor, an onboard computer can detect changes in pixel patterns which can help a drone hover nicely at low altitudes. Think of it like a mouse that tracks your desk to help the cursor move. There are obviously drawbacks to this technology. A patterned floor can confuse the flow algorithms to be able to track the ground properly. Secondly, the precise altitude from ground is required for this to work properly. Hence, sonar/ultrasonic sensors were added along the camera to help with altitude data. Pick up any modern, high-end drone today and underneath it you will find a camera and a sonar sensor. These two sensors combined provide tremendous flight robustness at LOW ALTITUDE. Yes, there is a drawback to this system and it’s the altitude. This system only works well up to a few meters from ground. Single beam LiDAR is the solution to that. It can give proximity data of up to 40m very accurately. This is a great technology to track the true height from ground of a drone. The benefits of LiDAR sensors become obvious when you are flying on uneven grounds. It is also a great technology for obstacle avoidance on drones. As drones become more available, people will want to use them in changing environments. Obstacle avoidance is a great step forward for drone technology, and LiDAR sensors are the backbone of this.
The last type of sensor I will talk about is RTK GNSS (author’s note: RTK = Real Time Kinematics, GNSS = Global Navigation Satellite System). Precision localisation is vital for industrial drone applications such as mapping. This technology combined with internal IMU data provides great meta-data for data collected by other sensors on a drone. The only drawback to this technology is the complexity and the un-reliability. Many companies are trying very hard to make them more robust.
In your opinion, what are the advantages of LiDAR for drone awareness applications?
LiDAR is an amazing piece of technology. This technology has come a long way, and now there are tiny LiDAR sensors with great range and power consumption. Most people don’t realize this, but the bread and butter of LiDAR is the field of view of the sensor. With a super narrow field of view (typically 3 degrees) , LiDAR sensors are capable of detecting small objects from very far. In my book, LiDAR has two key advantages for drones: used as a precise altimeter, this sensor can help with precision flights over uneven grounds. Mapping and surveying applications, especially in “precision” agriculture, require consistent data over a long time. Altitude consistency cannot be provided from the barometer on board all flight controllers. Atmospheric pressure and temperature changes can corrupt the data. LiDAR, on the other hand, provides precise and accurate data under all environments. One may wonder why altitude is so important if the images taken are being stitched in post anyway. Some height change is fine right ? Unfortunately not. The height of your drone provides the pixel/cm resolution of your visual data. Precision height data is also very useful for other drone applications such as atmosphere monitoring. I was approached by some scientists who wanted to put a spore sensor on a drone. They wanted to fly the drone at different altitudes to collect spore data. A similar application to this is drones for weather monitoring. These drones require precise altitude data as well.
The second great application of LiDAR is obstacle avoidance. I will explain this in a greater detail in a latter question.
Below: LeddarOne Pixhawk Integration by Rai Gohalwar
You recently reviewed the LeddarOne sensor. Can you sum up what you think of this sensor, and what are its main benefits compared to similar technologies and products?
LeddarOne is a unique sensor. I spent some time tinkering with it and also deployed it on my drone. The one thing that stands out to me is the attention to detail. LeddarTech has thought of everything when designing this sensor. There are some distinct features of this sensor that make it apart from other similar sensors out there. First of all the multiple output array of diodes helps increase robustness significantly. Most other LiDAR sensors have one out beam which is not as robust, comparatively. I covered all but one of them and the sensor was still working flawlessly. When deploying thousands of these on drones, it is important to have the reliability that this sensor provides.
The other great benefit of this sensor is the fact that it communicates on a serial TTL bus. This ensures accurate data which is relative as well. You see, similar LiDAR sensors out there have analog connectivity. This method of communication can introduce discrepancy in data and offsets. Often, those sensors need to be calibrated when connected to the drone and the offset needs to be adjusted. LeddarOne solves this problem and again makes this sensor easy to deploy in mass numbers.
The other great advantage of the LeddarOne is the configuration tools. The Windows GUI is very easy to use and very powerful as well. This sensor is easy to customize for the users requirements which is something you don’t see with other similar sensors out there.
What do you think of multi-segment LiDAR sensors for drones? What could be new applications made possible with higher resolution sensors, as opposed to single-point sensing?
Multi-segment LiDAR technology has great potential on drones. The fact that one sensor can detect multiple points without any moving parts still gives me goosebumps. The robustness aspect of this technology is the cherry on top of a already huge cake. Some would have to come out of cave to not realize the obvious and most practical application: obstacle avoidance. Using a LiDAR sensor with an array of sensing points can enhance a drone’s ability to avoid objects in its trajectory. Many people have tried using LiDAR sensors for this purpose on drones before but with a spinning/scanning sensor approach. This approach is a nightmare. Unless you are using a $20000 sensor, there will be issues with the moving parts. Also, the sensor will not be able to track all sides all the time. Using 4 multi-segment LiDAR sensors provides data from all sides, all the time without any mechanical limitations or maintenance/problems.
What is your opinion about the state of collision avoidance systems for drones ? Are they delivering the goods? Which are, in your opinion, the technological keys to true, reliable sense-and-avoid capabilities?
Collision avoidance is still long ways to being mature on drones. Companies like DJI have deployed collision avoidance on their drones, but custom integration is still not common. Unless there is seamless integration of 360 degree avoidance system to open source UAV platforms, this technology will only be deployed on consumer and very high-end industrial drones. Companies are deploying different types of avoidance technology, such as ultrasonic, laser, and camera. All of them have their pros and cons. Ultrasonic is reliable, so is LiDAR and camera is ok in good lighting conditions. In dark environments, 3d imaging for proximity doesn’t perform well. One key concept they all lack is diverse environmental data.
Consider this: most drones that advertise collision avoidance features have sensors pointing front, left and right. These sensors only give one channel proximity data. This data is ok if you only want to detect walls and stop the drone from running into them. For true and smart collision avoidance there needs to be multiple detection points in each axis. With some added computing on board, this data can help the drone navigate smartly through tough courses, creating a true collision avoidance system. Multi-segment LiDAR is the ideal solution for genuinely capable avoidance systems. LiDAR has the range, field of view and robustness required for an efficient avoidance system. I must emphasize, turnkey is the real game changer. Its gotta be at a point where you just plug in 4 multi-segment sensors to a processor and let it do its thing.
What is your prediction regarding the next decade for drones and UGVs? Where is the industry going, for recreational users as well as for industrial / utility perspectives?
Oh man, I can talk about this all day. Drones have grown exponential interest in the last few years. Let’s talk consumer first. Drones will soon be a household toy much like Lego or gaming consoles. Anything that flies has always been cool to play around with. I think there is still a big consumer market for drones. Consumer drones are at the dawn of mass-market penetration. The challenge is how do you control millions of these tiny buzzers who can fly very high. I think there will be strict regulations for recreation drone use, much like fireworks. As far as technology goes, these drones are already so easy to fly, have the latest and greatest technology and are very small and convenient to carry.
The real bread and butter however is on the industrial side. Drones are deemed to be a must-have tool for any industrial workplace. The industry is off to a shaky start because there are no set standards for this technology. Many small companies have tried doing that, but since their products are not cross-platform, its very hard to convince multiple industries of what they need. The industrial drone future is very bright. I think people are realizing that robustness at a bare stack level is very important for an industrial product and there will be products out there that are great for farmers, construction companies and other drone users. Tech wise, we will definitely see much more powerful computing power on these rugged drones with IoT functionality as well. This will help process collision avoidance data, other sensors data and increase robustness. My only concern is adaptability. There are endless applications for drones, but currently the drones being used (mainly DJI) are closed source. There is an amazing SDK, but custom sensor integration is very hard. A platform with a well-thought out API and easy integration to custom sensors will be a game changer on the industrial side.
I just want to say that LeddarTech is an awesome organization with extraordinary staff. I am grateful to have the opportunity to explore the sensors you guys offer and test them firsthand on a drone. My experience with LeddarOne was amazing. Its not often you run into well-thought out and well-engineered products. When I think about the future of sensors, whether that’s on drones or cars, I am confident that LeddarTech will have a big role in it.
Learn more about LeddarTech products and technology for drones: