Working Through The Math Behind Robot Grippers
The book by Murray, Li, and Sastry is dense. The solution manual helps, but using it correctly takes some understanding of how the material connects. Most people approach it the wrong way and waste weeks going in circles on chapters like contact mechanics or nonlinear control. It covers all the major problem sets from the textbook. The first half focuses on rigid body kinematics and dynamics - forward and inverse kinematics for serial manipulators, Jacobian analysis, and the geometry of configuration spaces. The second half moves into planning, manipulation, and control with contact forces. The solutions are thorough, showing intermediate steps rather than just final answers, which matters when you're trying to understand where a matrix trace goes wrong. I spent about three weeks debugging an inverse dynamics problem for a six-degree-of-freedom arm before I realized I'd been transposing the Jacobian in my derivation. The solution manual caught it in the equivalent problem set. Specifically, chapter 4 on Lagrangian dynamics has a note about the Coriolis terms that most people skip over, but getting that sign wrong cascades through every simulation you run afterward.
One thing the manual doesn't do well is explain the physical intuition behind the math. It will show you the proof for the constraint equations in chapter 8, but it won't tell you why a four-bar linkage constraint surface is non-convex in general configuration space. You have to cross-reference with lecture notes or additional papers for that gap.
How I Actually Used It Over Two Semesters
I didn't read it cover to cover. That's a mistake. I used it selectively after attempting each problem set on my own, then compared my approach against the solution. The biggest time sink is chapter 5 on robotic grasp stability - the positive real lemma conditions for form closure aren't straightforward, and I wasted maybe ten hours deriving them from scratch before checking the manual's approach using nullspace projections. For the differential kinematics sections, the manual's notation matches the book exactly, which saves you from translation errors. Some other resources use different angle conventions for Euler sequences, and that alone can throw off your entire Jacobian derivation. The manual sticks to the standard ZYZ convention throughout, so there's no confusion there. If you're working through the chapter on hybrid motion control, pay attention to how the solution decomposes the task space. The manual shows the decomposition explicitly using projection matrices, which is something you'll need for any real implementation. Simulating the hybrid controller without that decomposition is basically impossible because the force and motion constraints conflict at the boundary.
Get the Full Details
The main limitation is that the manual assumes you're comfortable with Lie groups and exponential coordinates. If you're not, chapters 2 and 3 will feel opaque even with the solutions. I'd recommend pairing it with video lectures or a supplementary text like Modern Robotics by Lynch and Park, which covers the same group theory material with more pedagogical scaffolding. There's also a gap in the later chapters where solutions skip over certain numerical methods. The chapter on dexterity measures gives the analytic formulas but doesn't walk through the computational cost or singular value thresholding needed when you evaluate them numerically. I had to write my own code to handle the conditioning issues when computing the condition number of the grasp map for underactuated hands.
Practical Workflow
Attempt the problems first. Don't skip this step. The retention is minimal if you just read solutions passively. When you're genuinely stuck on a specific step - say, the Lie algebra commutator expansion in problem 3.14 - look at only that part of the solution. Don't peek ahead. For the contact mechanics problems in chapter 8, I found it useful to verify each matrix dimension before proceeding, since the manual occasionally uses condensed notation that hides the actual sizes involved. The file organization is a bit messy if you're trying to search through it digitally. The chapter groupings are clear but the problem numbers jump around between editions. Make sure you have the edition that matches your textbook, because the pagination and problem ordering differ between the 1994 original and the 2017 reissue. For the nonlinear control sections, the manual's solutions assume you already know how to construct Lyapunov functions by hand. If you're still learning that skill, you'll find the solutions move too fast through the stability proofs. I found success by working backward from the final Lyapunov candidate in the solution to reconstruct the derivation steps, which took me about an hour per problem but clarified the whole approach.
