diff --git a/libraries/AP_Math/AP_Math.cpp b/libraries/AP_Math/AP_Math.cpp
index ddc8b3f900..41d2e8693d 100644
--- a/libraries/AP_Math/AP_Math.cpp
+++ b/libraries/AP_Math/AP_Math.cpp
@@ -82,6 +82,16 @@ template float safe_sqrt<short>(const short v);
 template float safe_sqrt<float>(const float v);
 template float safe_sqrt<double>(const double v);
 
+/*
+  replacement for std::swap() needed for STM32
+ */
+static void swap_float(float &f1, float &f2)
+{
+    float tmp = f1;
+    f1 = f2;
+    f2 = tmp;
+}
+
 /*
  * linear interpolation based on a variable in a range
  */
@@ -89,6 +99,11 @@ float linear_interpolate(float low_output, float high_output,
                          float var_value,
                          float var_low, float var_high)
 {
+    if (var_low > var_high) {
+        // support either polarity
+        swap_float(var_low, var_high);
+        swap_float(low_output, high_output);
+    }
     if (var_value <= var_low) {
         return low_output;
     }
diff --git a/libraries/AP_Math/AP_Math.h b/libraries/AP_Math/AP_Math.h
index bd5ad2292a..7af6e999bd 100644
--- a/libraries/AP_Math/AP_Math.h
+++ b/libraries/AP_Math/AP_Math.h
@@ -248,6 +248,9 @@ inline constexpr uint32_t usec_to_hz(uint32_t usec)
 
 /*
   linear interpolation based on a variable in a range
+  return value will be in the range [var_low,var_high]
+
+  Either polarity is supported, so var_low can be higher than var_high
  */
 float linear_interpolate(float low_output, float high_output,
                          float var_value,