UAV designs are a perpetual compromise between the ability to fly long distances efficiently with payloads (fixed-wing) and the ability to maneuver, hover, and land easily (rotorcraft). With a very few rather bizarre exceptions, any aircraft that try to offer the best of both worlds end up relatively complicated, inefficient, and expensive. The ideal fantasy UAV would be a fixed-wing aircraft with the magical ability to land on a dime, and a group of researchers from the University of Sherbrooke in Canada have come very close to making that happen, with a little airplane that uses legs and claws to reliably perch on walls.
The majority of the perching robots that we've seen are quadrotors. Perching with a quadrotor is significantly easier than perching with a fixed-wing aircraft, because you have many more degrees of control, and you're not obligated to keep the vehicle moving forward all the time. There are certainly quadrotors that take more creative approaches to perching, but it's also worth noting that perching is less valuable to quadrotors, since they can hover and land relatively easily.
Fixed-wing perching is a more difficult problem, since perching generally requires the vehicle to be very close to stationary to keep it from just bouncing off of whatever surface it's trying to land on. Airplanes are not happy with being stationary in mid air, and will usually stall as soon as air stops moving over their wings, because that's where the majority of their lift comes from. If you're very (very) careful, you can time this stall to coincide exactly with your perch target: that is how MIT's powerline perching glider works.
Dino Mehanovic, John Bass, Thomas Courteau, David Rancourt, and Alexis Lussier Desbiens from the University of Sherbrooke realized that perching with a fixed-wing aircraft doesn't need to involve a stall to achieve that vertical and ultra low-speed approach, as long as you can maintain control over the aircraft. Birds do this all the time, in fact, by using the thrust from their wings to controllably approach objects slow enough to enable a comfortable perch. Sherbrooke's Multimodal Autonomous Drone (S-MAD) uses a similar thrust-assisted landing technique (along with some microspine feet) to reliably perch on walls, and then take off again:
There are several tricks to this. The first trick is the pitch-up maneuver, which turns the fixed-wing airplane into a temporary helicopter of sorts, relying entirely on the propellor for lift generation (the thrust to weight ratio is 1.5) while the wings provide enough of a control surface to cancel out the torque. At that point, the UAV can approach the wall as slowly as you like (using a laser rangefinder for wall detection), which leads to the second trick: maximizing the “zone of suitable touchdown conditions,” or making sure that the approach is slow enough and steady enough that you can perch reliably with little hardware (sensing and otherwise). And the third trick is having a perching system, legs and microspines in this case, that are flexible enough to achieve a robust perch even if the aircraft isn't doing exactly what you'd like it to be doing.
In indoor testing, S-MAD went 20 for 20 in successful perching experiments, which is pretty good even though the environment was (we assume) very tightly controlled. Future work will include adding some sensors to help with the final phase of wall contact, and also working on thrust-assisted wall climbing, managing aborted approaches, and recovering from perch attempts where the microspines don't grab onto the wall properly. If they can put all of this together, and get it to work robustly outdoors on different surfaces and under different conditions, S-MAD (or UAVs like it) could become the go-to systems whenever you need an efficient and reliable long-range, long-duration platform that can also pull over onto the nearest wall or tree trunk from time to time.
For more details, we spoke with Alexis Lussier Desbiens via email.
IEEE Spectrum: Can you explain what makes S-MAD unique among perching robots (or winged perching robots specifically) and why the capabilities that it offers are important?
Alexis Lussier Desbiens: It is true that there are now quite a few great quadcopters that can perch on various surfaces, smooth or rough. We still like fixed-wing though. They are particularly efficient in flight, and could move much faster between perching sites.
We know of only two other fixed-wing perching airframes: the sticky-pad plane (Anderson et al., 2009) and the dart mechanism (Kovac et al., 2009). Both of these systems work well at small scales. However, as scale is increased, the flight speed also increases, which makes dead-on impact tricky. With higher approach velocities, the force and acceleration required at impact to dissipate the momentum in a reasonable distance (i.e., the suspension travel) are also larger, which is why birds aerodynamically shed a good part of their speed before touchdown. That is what we are trying to reproduce with this platform. Even with a relatively heavy platform, we can shed a significant part of the airplane speed by using the pitch up maneuver and thrust to create a smooth landing with reasonable forces.
I started this work at Stanford on a glider. I was able to demonstrate landing on a fairly light platform and takeoff on a different platform that was much heavier. It was extremely difficult and unreliable to do both landing and takeoff with the same platform. That is what Dino has achieved, and what we are describing in this paper.
How reliable is the perching right now? What kinds of failure recovery techniques are you working on?
It is reliable in the lab. We performed more than 20 successive landings from different approach speeds. More is required, but after that we got bored! Through simulations, we designed the system to be robust to variations in numerous parameters. So far, we performed about a dozen landings outside on calm days. We want to keep working outside to push the limits of our system.
We are thinking about various failure causes (unsuitable states during the approach, smooth surface for the microspines) and failure detection timing (before touchdown, at touchdown and after touchdown). In all cases, the takeoff strategy allows us to abort the maneuver at different stages, increase the thrust, and fly away from the wall to try again or find a different site.
How would your perching approach change if you were to use a different kind of surface engagement hardware?
We like microspines because they are so simple and light. However, they have pretty strict force constraints that need to be respected for successful adhesion. That is why a suspension is required for semi-passive high speed landings. On smooth surfaces, directional dry-adhesives (e.g., gecko inspired) could be used. As they have similar force constraints, they could be use with few modifications.
Cutkosky et al. developed many “opposed grippers” in the last few years based on microspines and dry-adhesives. Due to the gripping internal forces, these have a larger operating force space. They are not as sensitive to rebound and work well for perching, as they demonstrated on quadcopter. With this kind of mechanism, it would be even easier to land and remain on the surface.
Other technologies do exists: magnetics, electroadhesion, etc. In all cases, our final approach toward the wall is now fairly smooth, controllable and at low speed. It shouldn’t be a problem to integrate most of these technologies.
Can you describe the thrust-assisted climbing process that you're working on? When would it be more advantageous to use this rather than taking off and perching again?
We are looking into that right now, both how to do it and when it makes sense to use various modes. In its most simple incarnation, thrust assisted climbing would consist of turning on the propeller, allowing the microspines to slide up while remaining into contact and turning the propeller off. This is enabled by the microspines anisotropic sliding behaviour: they slide in one direction and catch in the other. We could also think of flying up some distance away from the wall, or integrating actuated legs.
Compared to climbing, flying up would allow the drone to move fast. Propellers can also be fairly efficient. However, trade studies to compare different locomotion modes are always tricky. You have to define your objectives: cost of transport, speed, agility, etc. You also have to consider numerous factors that are sometime hard to quantify: efficiency of gears, reuse of some components between flight and climbing, transition time, propeller size, operating away from the design point, battery size, etc. Pope et al. (2016) have done measurements and estimations on their SCAMP platform. Flying seems to be a more efficient way to climb up one meter, but they claim that motor/gearbox efficiency may be improved sufficiently to reverse that. Compared to SCAMP, our platform would spend significantly less time in transitions as the propeller is already pointing up.
What kinds of practical applications could a vehicle like this be used for?
This kind of capabilities enables small UAVs to perform extended mission, for days or weeks, where you land, rest/recharge, takeoff and repeat ad infinitum. This allows new types of missions. Ultimately, such platform could be used for long duration surveillance, energy harvesting, inspection of structures or reconfigurable sensor networks.
Autonomous Thrust-Assisted Perching of a Fixed-Wing UAV on Vertical Surfaces, by Dino Mehanovic, John Bass, Thomas Courteau, David Rancourt, and Alexis Lussier Desbiens from the University of Sherbrooke in Canada, was presented at the 2017 Living Machines Conference at Stanford, where it won a Best Paper award.
Evan Ackerman is a senior editor at IEEE Spectrum. Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.