Play with two robots each. One of the premade courses in the rules (Interference, on page 26 in this edition) actually suggests this as a variation, where you have one "racer" robot that can touch flags, and one "blocker" that can't. That way the board ends up more crowded and you have an incentive to actually go after the other player's robots rather than just heading for flags. (The two robots on a team are distinct, with separate hands and separate programs.) This should work reasonably well with a variety of maps, not just the one it's suggested on.
I think that one is the most straightforward for two players, and directly encourages interaction and damage, but the team variants do include the other obvious possibilities with a two-robot team:
- from "Tandem Carnage" (page 29) - either robot from the team can touch each flag, as long as they're overall in order.
- from "All for One or One for All" (page 30) - one robot from the team must touch all the flags in order, but it can be either one.
For what it's worth, there are also some more elaborate team games that could work with one player controlling the whole team, but they probably change the game more than you wanted, and require a lot more care in map construction:
- "Capture the Flag" (page 31) - what it sounds like! Two boards, pushing a robot and surviving the phase counts as a "tag" and sends the other robot back to its home board, you can pick up the enemy flag and carry it, etc.
- "Toggle Boggle" (page 32) - touch all the flags, but only the last team to touch a flag counts.
- "War Zone" (page 32) - just try to run the other team out of life tokens.
I'm not a huge fan of trying to amp up the impact of damage (whether by making the track smaller or making each hit count more), because the core problem is simply that you don't run into each other that much. So yes, the couple of stray hits will matter, but they'll still be rare and if they do matter it'll feel more random, like you got punished a lot for a small mistake.
Reducing the board size as Toon Krijthe suggests is also a good idea, though there's a tradeoff that might matter to some people. It will indeed force the robots together, but it'll tend to favor smaller-scale movement (and more rotation), and have less of the larger-scale movement, multiple choices of paths, and diversity of board elements you get on larger boards.