The mechanical work is basically complete! I've been running in to trouble with the inverter. Reliable motor control is evading me, as is effective hardware protection. I've blown a few inverters at this point and am taking a step back to reconsider my plan for this final (and most important) piece of the puzzle.
Instead of continuing forward with my own control board, I've decided to try my hand at reverse engineering the communications between the stock inverter and the hybrid controller. I figure that the engineers at Lexus/Toyota, with all the resources and expertise at their disposal, will have tuned the motor control far better that I could ever hope to in my garage. The hardware protections would also be as designed so the risk of blowing an inverter should be much lower than with my amateur attempts at stable IPM motor control. The only way for this project to be reliable enough to act as my daily driver with the occasional thrashing is to leverage the effort already put in to the powertrain in Japan.
To that end, I spend last weekend with a borrowed 2007 Camry Hybrid. This car uses a very similar inverter (same control board inside, right down to the part number!), so I set up my Saleae logic analyzer and recorded the communication between the inverter and hybrid controller while performing a number of maneuvers and taking careful notes.
There is a LOT of data to sift through, but I've managed to isolate the pieces of data which contain the torque commands for each MG, as well as some parts which look like "modes" or "states". I will spend some more time this coming weekend with one of my spare inverters to see if I can manage to get the hardware to respond to these commands. Hopefully I can get away with feeding it the same data I recorded, only varying the torque commands, and have it respond.