Caltech’s Cannon-Launched SQUID Drone Doubles in Size, Goes Autonomous

Foldable, ballistically launched SQUID drones could one day deploy to Mars

5 min read

Evan Ackerman is IEEE Spectrum’s robotics editor.

SQUID drone
Cannon-launching the SQUID drone: Inside the launcher tube [left], deploying the arms and fins [center], and fully deployed configuration [right]. The pictures are at intervals of 82 milliseconds.
Image: Caltech & NASA JPL

At IROS last year, Caltech and NASA’s Jet Propulsion Lab presented a prototype for a ballistically launched quadrotor—once folded up into a sort of football shape with fins, the drone is stuffed into a tube and then fired straight up with a blast of compressed CO2, at which point it unfolds itself, stabilizes, and then flies off. It’s been about half a year, and the prototype has been scaled up in both size and capability, now with a half-dozen rotors and full onboard autonomy that can (barely) squeeze into a 6-inch tube.

SQUID stands for Streamlined Quick Unfolding Investigation Drone. The original 3-inch (7.6-centimeter) SQUID that we wrote about last year has been demoted to “micro-SQUID,” and the new SQUID is this much beefier 6-inch version. You should read our earlier article on micro-SQUID for some background on this concept, but generally, tube-launched drones are unique in that they remove the requirement for the kind of specific takeoff conditions that most drones expect—stationary and on the ground and not close to anything that objects to being sliced to bits. A demonstration last year showed micro-SQUID launching from a moving vehicle, but the overall idea is that you can launch a SQUID instantly and from pretty much anywhere.

The point of micro-SQUID was to work out the general aerodynamic and structural principles for a ballistically launched multirotor, rather than to develop something mission capable. Mission capable means, among other things, onboard autonomy without reliance on GPS, which in turn calls for sensing and computing that’s heavy and power hungry enough that the entire vehicle needed to be scaled up. The new 6-inch SQUID features some major updates, including an aerodynamic redesign for improved passive stabilization during launch and ballistic flight through the use of deployable fins. The autonomy hardware consists of a camera (FLIR Chameleon3), rangefinder (TeraRanger Evo 60m), IMU/barometer (VectorNav VN-100), and onboard computer (NVIDIA Jetson TX2).

Caltech SQUID canon drone

Caltech SQUID canon droneTop: SQUID overview. Bottom: SQUID partially inside the launcher tube (a), with its arms and fins fully deployed from a side (b), and top perspective (c).Image: Caltech & NASA JPL

The structural and aerodynamic changes are necessary because SQUID spends the first phase of its flight not really flying at all, but rather just following the ballistic trajectory that it’s on once it leaves the launcher. If it’s just going straight up, that’s not too bad, but things start to get more complicated if the drone gets launched at an angle, or from a moving vehicle. Having a high center of mass helps (the battery lives in the nose cone), and deployable fins pull double duty by keeping the drone passively pointed into the airstream while also serving as landing gear—without the fins, it would start to tumble after leaving the tube, and then good luck trying to control it. In order for the fins to be both foldable and stable enough for SQUID to land on, they’ve got a latching mechanism that helps keep them rigid, and apparently once everything got put together it took a little bit of sanding of the arm hinges before the drone would actually fit into the launch tube.

That 6-inch hard stop on the diameter of SQUID turned out to be a real challenge. Most drones are power or mass constrained, but SQUID is instead volume constrained. Not only do you have to cram all of your batteries and computers into that space, you have to make sure that the sensors have the field of view that they need while keeping in mind that in its folded state all the arms and legs have to share the same space as everything else. It turns out that SQUID is very well optimized, though, weighing just 3.3 kilograms, only about 0.3 kg more than what the roboticists estimate a nonfoldable, nonoptimized conventional drone with similar capabilities would weigh. 

So why bother with all of this hassle for the whole tube launch thing? There are a bunch of reasons that make it worth the effort: 

  • It’s fast to launch. There’s no unpacking or setting up or finding a flat spot or telling everyone to stand back, just push a button and bam, SQUID is out of the tube at 12 meters per second and in flight. 
  • It’s safe to launch. Unless someone is sitting directly on top of the launch tube (in which case you could argue that they deserve what they’re about to experience), the launch rapidly clears human level before deploying any dangerously spinny bits. 
  • It can launch while moving. This is a big one—the ballistic launch and self-stabilization means that SQUID can be reliably launched from a moving vehicle moving at up to 80 kilometers per hour, like a truck or a boat, significantly increasing its utility, especially in emergency scenarios.
  • It can sometimes launch through things. The researchers point out that in its most aerodynamic shape (without fins or rotors deployed), SQUID could potentially be launched straight through tree canopies or power lines if necessary, which is a totally unique capability for a rotorcraft.

We asked the researchers about their experience developing the larger version of SQUID, and they shared this behind-the-scenes story with us about how they managed to set things up so that they didn’t crash even once:

Moving to a larger SQUID was hard technically (as we had to design an entirely new vehicle), but the testing logistics was a huge jump in difficulty. For our smaller SQUID, simply a net and some spare parts would suffice to keep testing going for a day. But when we moved to the bigger SQUID, we’re throwing something a lot heavier, and packed with expensive electronics for autonomy, into the sky. 

An indoor tether system was challenging to set up because the height of the CAST arena (42 foot-tall) meant the ideal locating point for the tether was completely inaccessible without a cherry-picker. The Caltech Drone Club stepped up, and helped construct the tether system by weaving a tiny quadrotor towing fishing line around the ceiling beams. The fishing line was then used to pull larger ropes through.

One of the interesting things that was learnt with the tether system was the extreme acceleration of SQUID as it exited the launch tube meant the tether cable becomes very slack and actually risks getting tangled with or cut by the propellers. Luckily our incremental testing campaign caught this before we had any incidents. To deal with this slack tether situation, we constructed a nose cone with a 5 foot carbon fiber tube mounted at the apex, which we called SQUID’s swordfish nose (we had a bit of an aquatic theme going already). A tether attached to SQUID’s frame runs through the tube and connects to the larger CAST tether system. We confirmed that during launch (for our given launch parameters), the tether never droops lower than the tube, so we prevented all tether-propeller interactions.

As you might expect from a drone from Caltech and JPL, long term the plan is to start thinking about aerial deployment—like, launching small drones from larger aircraft. This could eventually provide a way for small drones to be deployed from spacecraft on Mars during atmospheric entry, potentially reducing the need for a large lander. In fact, it’s common for aeroshells that deliver landers to planetary surfaces to rebalance themselves during atmospheric entry by dropping a bunch (like, 150 kg) of weight to adjust their angle of attack. Those weights are utterly useless chunks of tungsten, but if it was possible to drop some midair-deployable drones instead, you could potentially do a whole lot of extra science without adding extra mass or risk to an existing mission.

Design and Autonomous Stabilization of a Ballistically-Launched Multirotor,” by Amanda Bouman, Paul Nadan, Matthew Anderson, Daniel Pastor, Jacob Izraelevitz, Joel Burdick, and Brett Kennedy from Caltech and JPL, was presented at ICRA 2020, where it was awarded best paper in Unmanned Aerial Vehicles.

The Conversation (0)