Training the APOPO HeroRATs

Habituation – the rats are “socialised” (handled) when they are young and still living with their mothers. Uncle Albert is responsible for the breeding programme and for habituation. The habituation phase introduces the baby rats to different people and environments, including walking on different surfaces and hearing different sounds (including loud noises – some of the rats will go on to become HeroRATS and they need to be “bomb-proof” when it comes to loud noises around them).

Clicker – a very well-known reward-based animal training method. Animals work hard to get food rewards (known as primary reinforcers), but it’s tricky to time the food reward perfectly to enable it to be used to pin-point exact behaviour that you want to reward. This is where a clicker comes in because it can be accurately timed to coincide with the action you want to reward. Firstly, you have to build an association between the animal hearing a clicking sound and then receiving the food reward. Next, the clicker is turned into a bridge between an action/behaviour and a reward. Actions might include a rat scratching the ground when it detects a certain odour, or a dog sitting when it hears the command “sit”. The rats generally take to the clicker training very quickly, and within a week they’ll be trotting towards the trainer to get a reward when they hear the sound. At APOPO, this is where the rats are introduced to the soil table – a raised platform covered with a thin layer of Tanzanian soil so that they can get used to the feel of running across it. The rats are also placed in cute little hand-made harnesses so that they can run along a piece of string between the two trainers at opposite ends of the table.

Indication – this is where the rats are introduced to the tea eggs. These holey metal spheres are placed on the soil table for the rats to discover. The rats run up and down the table in a little search pattern and when they come to a tea egg, if they stop and sniff the egg, they get a clicker and a reward – this is how they learn that they have to look for the tea eggs. During this phase they will learn to signal more strongly every time they encounter a tea egg.

Discrimination – this is where the target scent is introduced into the some of the hollow tea eggs on the soil table. This scent is TNT for the HeroRATS-in-training, but it can be other things if the rats are being trained to participate in experiments. This is a tough stage – the rats only hear the clicker when they stop and sniff the TNT-filled eggs and they must learn to ignore the ones that don’t smell of TNT. They’ve got to get this stage right as it paves the way for the big next step.

Soil floor – the rats leave the small soil table and begin working with their trainers in a purpose designed soil-filled pit that is more representative of the minefield environment. They still work with TNT-filled and empty tea eggs and get rewarded for identifying and signalling when they find a TNT-filled egg. During this stage the trainers begin to bury the tea eggs in the soil so that the rats have to scratch around to get them. The rats are now big enough to carry the tea eggs and in the excitement of hearing the clicker, lots of the rats learn to run with the egg over to the trainer. It’s adorable! 

The Mine Detection Rats then graduate to Basic Field Training and the Tuberculosis Detection Rats are introduced to their work environment for further training.

Tea Egg Area – the rats are trained in a new environment – in a minefield outdoors, a few miles from APOPO Headquarters – but they begin with the familiar tea eggs. The floor is similar to the soil floor area and the rats get used to searching larger areas in one go, shuttling back and forth between the trainers. This is a good stepping stone to get the rats used to the wide-open space in the minefield training area.

3 metre minefield – this training stage is the first time the rats are exposed to real landmines, and it’s the first time they’ve run around with grass tickling their tummies! These landmines have been deactivated (so they can no longer explode) and have been buried for the last 20 years, so it’s fairly representative of a real minefield – there is no disturbed soil to give rats a clue about where the mines are buried. There are many different 3-metre wide minefield strips for the rats to train on, so they can’t learn the location of the mines in one area with practice. Along the edge of the strip there are markers to explain to the trainers where the mines are so that they know when to reward a signalling rat. The rats develop stamina here as they are not used to shuttling back and forth for long – it could take them 20 minutes to clear a small area at first, but they get quicker with practice.

5 metre minefield – this stage builds stamina from the last stage. The strips are now 5 metres wide and many of the strips are 32 metres long, so the rats have to make lots of trips to clear the whole box in 20 minutes. The trainers each take a step further down the strip each time a rat completes one journey across the gap between them, and the rats can sense a mine without going directly over it, so the rats and trainers are able to make good progress down the strip. Some of the boxes don’t have any mines in them at all – the rats need to get used to a little disappointment because out on their future operations, landmines are relatively rarely encountered when clearing ground. So, there are still treat rewards, but fewer than in the early phases of training.

When the trainers believe the rats are ready, they take a blind test, where there are no markers to show the trainers where the mines are buried, and an assessor marks on a map each location that the rat signals and only after the maps have been checked against the database can the trainers see if the rat signalled correctly at 100% of the buried mines. It is acceptable for a rat to sometimes signal when there isn’t a mine buried there, but it’s absolutely essential that they never miss a mine, as that could lead to devastating consequences. 

Once a rat can pass the blind tests for the Basic Minefield Training then it can graduate to Advanced Training

Advanced – here the rats clear two 5-metre boxes per 20-minute session. There are often no mines in a box and only some of the buried mines are marked, so the rat’s signals have to be noted down and checked against the database to see how it is performing. This phase involves quantity – the rats clear two boxes every day for around 2 months to get their practice in before they are ready to graduate as a fully qualified HeroRAT that is ready to go out and give people their land back, free from landmines. 

One point on blind-testing of mines – it’s inevitable that sometimes the trainer’s behaviour will unconsciously signal to the rat that it’s approaching a mine, whether it’s that they unconsciously fidget with the clicker, hold their breath or suddenly go silent and pay close attention. Having a significant amount of blind training thrown in allows the trainers to be sure that the rat is capable of detecting mines with absolutely no input from the trainer.

It is perfectly possible for a rat to fail the test at the end of any stage and have to continue on training at the same stage, or return to a previous stage, or it’s possible that the training team will decide that a rat just hasn’t quite got what it takes to become a fully-fledged HeroRAT. There are plenty of other avenues at APOPO, with lots of interesting detection projects in the pipe-line and even some great STEM engagement activities with kids.