Matlab代码报错‘Inner matrix dimensions must agree’排查求助

Let's break down why you're hitting that Inner matrix dimensions must agree error and fix it step by step:

First, Identify the Core Issue #

The error occurs at V1 = (R2-f*R1)/g; and V2 = (gdot*R1-R1)/g; because either:

• Your input vectors R1/R2 have mismatched row/column dimensions,
• The variables f, g, or gdot are accidentally matrices instead of scalars (as they should be in Lambert problem calculations),
• There's a typo in your V2 calculation (more on that later).

Step 1: Fix Input Vector Dimensions #

MATLAB is strict about row vs column vectors. If R1 is a 1x3 row vector and R2 is a 3x1 column vector, subtracting f*R1 from R2 will fail immediately.

Add this at the start of your function to force consistent column vectors:

function [V1, V2] = lambert(R1, R2, t, string)
    % Force input vectors to be column vectors
    R1 = R1(:);
    R2 = R2(:);
    r1 = norm(R1);
    r2 = norm(R2);
    % Rest of your code...

Step 2: Ensure Scalar Values for A, mu, and Helper Functions #

Your code is missing critical definitions that are likely causing non-scalar outputs:

1. Define mu as a scalar: This is the gravitational constant (e.g., 3.986e14 m³/s² for Earth). Add it at the top:

   mu = 3.986e14; % Replace with your target body's gravitational constant
   

2. Fix A definition: In Lambert problems, A is a scalar related to the cross product of R1 and R2. Define it properly:

   A = norm(cross(R1, R2)); % Scalar magnitude of the angular momentum term
   

3. Verify helper functions return scalars: Your S(z), C(z), F(z,t), and dFdz(z) functions must accept a scalar z and return a scalar. For example, here's the correct scalar implementation of Stumpff functions (required for Lambert calculations):

   function s = S(z)
       if z > 0
           s = (sqrt(z) - sin(sqrt(z)))/(sqrt(z))^3;
       elseif z < 0
           s = (sinh(sqrt(-z)) - sqrt(-z))/(sqrt(-z))^3;
       else
           s = 1/6;
       end
   end
   
   function c = C(z)
       if z > 0
           c = (1 - cos(sqrt(z)))/z;
       elseif z < 0
           c = (cosh(sqrt(-z)) - 1)/(-z);
       else
           c = 1/2;
       end
   end
   

   If these functions return matrices (e.g., if z accidentally becomes a matrix during iteration), y(z) will be a matrix, turning f, g, and gdot into matrices too—causing the dimension mismatch.

Step 3: Fix the Typo in V2 Calculation #

Your current V2 line has a clear mistake:

V2 = (gdot*R1-R1)/g;

This should use R2 instead of the second R1 (per standard Lambert velocity formulas):

V2 = (gdot*R2 - R1)/g;

Even if dimensions worked, the original line would produce incorrect velocity results.

Step 4: Validate Iteration Logic #

Double-check that z remains a scalar throughout your loops. You initialize it as -100 (a scalar), so as long as F(z,t) and dFdz(z) return scalars, z will stay a scalar—keeping y(z), f, g, and gdot as scalars.

Final Check #

Test with a simple, known Lambert problem case to verify:

% Example input: Circular orbit transfer
R1 = [7000e3; 0; 0];
R2 = [0; 7000e3; 0];
t = 5000; % Transfer time in seconds
[V1, V2] = lambert(R1, R2, t, '');

This should resolve the dimension mismatch error and produce valid velocity vectors.

内容的提问来源于stack exchange，提问作者martin